s390x: STFLE nested interpretation
Subsystems (4)
cover s390x: STFLE nested interpretation
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:
* simple pseudo random number generation (arch independent)
* exit (optionally with return code) from snippet (s390x)
Nina Schoetterl-Glausch (5):
lib: Add pseudo random functions
s390x: lib: Remove double include
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/libcflat.h | 7 ++
lib/s390x/sie.h | 1 +
lib/s390x/snippet-guest.h | 26 +++++
lib/s390x/{snippet.h => snippet-host.h} | 9 +-
lib/rand.c | 19 ++++
lib/s390x/sie.c | 29 +++++-
lib/s390x/snippet-host.c | 40 +++++++
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 | 12 +--
s390x/snippets/c/sie-dat.c | 19 +---
s390x/snippets/c/stfle.c | 26 +++++
s390x/spec_ex-sie.c | 2 +-
s390x/stfle-sie.c | 132 ++++++++++++++++++++++++
s390x/uv-host.c | 2 +-
22 files changed, 322 insertions(+), 39 deletions(-)
create mode 100644 lib/s390x/snippet-guest.h
rename lib/s390x/{snippet.h => snippet-host.h} (93%)
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.cbase-commit: 6b31aa76a038bb56b144825f55301b2ab64c02e9
Signature · 1 line
2.41.0
PATCH 1/5 lib: Add pseudo random functions
Add functions for generating pseudo random 32 and 64 bit values. The implementation is very simple and the randomness likely not of high quality.
--- Makefile | 1 + lib/libcflat.h | 7 +++++++ lib/rand.c | 19 +++++++++++++++++++ 3 files changed, 27 insertions(+) create mode 100644 lib/rand.c
Makefile +1 -0
@@ -28,6 +28,7 @@ cflatobjs := \ lib/printf.o \ lib/string.o \ lib/abort.o \ + lib/rand.o \ lib/report.o \ lib/stack.o
lib/libcflat.h +7 -0
@@ -83,6 +83,13 @@ extern void abort(void) __attribute__((noreturn));extern long atol (const char *ptr);extern char *getenv (const char *name); +typedef struct { +uint32_t val; +}rand_state ; +#define RAND_STATE_INIT (x) ((rand_state){ .val = (x) }) +uint32_t rand32 (rand_state *state); +uint64_t rand64 (rand_state *state); +extern int printf (const char *fmt, ...) __attribute__((format (printf,1 ,2 )));extern int snprintf (char *buf,int size,const char *fmt, ...)
lib/rand.c new file +19 -0
@@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* pseudo random functions +* +* Copyright IBM Corp. 2023 +*/ + +#include "libcflat.h" + +uint32_t rand32 (rand_state *state) +{ + state->val =0x915f77f5 * state->val +1 ; +return state->val ^ (state->val >>16 ); +} + +uint64_t rand64 (rand_state *state) +{ +return (uint64_t )rand32 (state) <<32 |rand32 (state); +}
Signature · 1 line
2.41.0
Re: [kvm-unit-tests PATCH 1/5] lib: Add pseudo random functions
Quoted 70 lines · Nina Schoetterl-Glausch
Add functions for generating pseudo random 32 and 64 bit values. The implementation is very simple and the randomness likely not of high quality.
Signed-off-by Nina Schoetterl-Glausch <[email protected]>--- Makefile | 1 + lib/libcflat.h | 7 +++++++ lib/rand.c | 19 +++++++++++++++++++ 3 files changed, 27 insertions(+) create mode 100644 lib/rand.cMakefile +1
-0@@ -28,6 +28,7 @@ cflatobjs := \ lib/printf.o \ lib/string.o \ lib/abort.o \ + lib/rand.o \ lib/report.o \ lib/stack.olib/libcflat.h +7
-0@@ -83,6 +83,13 @@ extern void abort(void) __attribute__((noreturn));extern long atol (const char *ptr);extern char *getenv (const char *name); +typedef struct { +uint32_t val; +}rand_state ; +#define RAND_STATE_INIT (x) ((rand_state){ .val = (x) }) +uint32_t rand32 (rand_state *state); +uint64_t rand64 (rand_state *state); +extern int printf (const char *fmt, ...) __attribute__((format (printf,1 ,2 )));extern int snprintf (char *buf,int size,const char *fmt, ...)lib/rand.c new file +19
-0@@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* pseudo random functions +* +* Copyright IBM Corp. 2023 +*/ + +#include "libcflat.h" + +uint32_t rand32 (rand_state *state) +{ + state->val =0x915f77f5 * state->val +1 ; +return state->val ^ (state->val >>16 ); +} + +uint64_t rand64 (rand_state *state) +{ +return (uint64_t )rand32 (state) <<32 |rand32 (state); +}Signature · 1 line
2.41.0
Alex Bennée posted a prng patch a long time ago that never got merged.
https://www.spinics.net/lists/kvm-arm/msg50921.html
would it be better to merge that?
Thanks,
drew
Re: [kvm-unit-tests PATCH 1/5] lib: Add pseudo random functions
Quoted 4 lines · Andrew Jones
Quoted 3 lines · Nina Schoetterl-Glausch
Add functions for generating pseudo random 32 and 64 bit values. The implementation is very simple and the randomness likely not of high quality.
[...]
Quoted 5 lines
Alex Bennée posted a prng patch a long time ago that never got merged.
https://www.spinics.net/lists/kvm-arm/msg50921.html
would it be better to merge that?
Well, it's hard to say what metric to apply here.
How good does the randomness need to be?
I chose a minimal interface that should be amendable to evolution. That's why the state is hidden behind a typedef.
I think this would be good for Alex' patch, too, and there is nothing gained by exposing the implementation by prefixing everything with isaac. My patch is also simpler to review.
But I'm certainly not opposed to better algorithms.
Quoted 3 lines
Thanks,
drew
Re: [kvm-unit-tests PATCH 1/5] lib: Add pseudo random functions
Quoted 8 lines · Nina Schoetterl-Glausch
Quoted 4 lines · Andrew Jones
Quoted 3 lines · Nina Schoetterl-Glausch
Add functions for generating pseudo random 32 and 64 bit values. The implementation is very simple and the randomness likely not of high quality.
[...]
Quoted 8 lines
Quoted 5 lines
Alex Bennée posted a prng patch a long time ago that never got merged.
https://www.spinics.net/lists/kvm-arm/msg50921.html
would it be better to merge that?
Well, it's hard to say what metric to apply here.
How good does the randomness need to be?
Better randomness would improve coverage for random sample set type tests and possibly help stress more for stress tests.
Quoted 4 lines
I chose a minimal interface that should be amendable to evolution. That's why the state is hidden behind a typedef.
I think this would be good for Alex' patch, too, and there is nothing gained by exposing the implementation by prefixing everything with isaac.
I agree.
Quoted 2 lines
My patch is also simpler to review.
But I'm certainly not opposed to better algorithms.
Do you mind reposting Alex's patch, maintaining his authorship, but with your interface changes?
Thanks,
drew
PATCH 2/5 s390x: lib: Remove double include
In reply to Nina Schoetterl-Glauschlibcflat.h was included twice.
--- lib/s390x/sie.c | 1 - 1 file changed, 1 deletion(-)
lib/s390x/sie.c +0 -1
@@ -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
2.41.0
Re: [kvm-unit-tests PATCH 2/5] s390x: lib: Remove double include
Quoted 3 lines · On Wed, 13 Dec 2023 13:49:39 +0100 Nina Schoetterl-Glausch <[email protected]>…
libcflat.hwas included twice.Signed-off-by Nina Schoetterl-Glausch <[email protected]>
Quoted 16 lines
--- lib/s390x/sie.c | 1 - 1 file changed, 1 deletion(-)lib/s390x/sie.c +0
-1@@ -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>
PATCH 3/5 s390x: Add library functions for exiting from snippet
In reply to Nina Schoetterl-GlauschIt 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 the host's.
---
s390x/Makefile | 1 +
lib/s390x/asm/arch_def.h | 13 ++++++++
lib/s390x/sie.h | 1 +
lib/s390x/snippet-guest.h | 26 ++++++++++++++++
lib/s390x/{snippet.h => snippet-host.h} | 9 ++++--
lib/s390x/sie.c | 28 +++++++++++++++++
lib/s390x/snippet-host.c | 40 +++++++++++++++++++++++++
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 +-
15 files changed, 123 insertions(+), 11 deletions(-)
create mode 100644 lib/s390x/snippet-guest.h
rename lib/s390x/{snippet.h => snippet-host.h} (93%)
create mode 100644 lib/s390x/snippet-host.cs390x/Makefile +1 -0
@@ -109,6 +109,7 @@ 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
@@ -504,4 +504,17 @@ static inline uint32_t get_prefix(void) return current_prefix; } +static inline void diag44 (void ) +{ + asmvolatile ("diag 0,0,0x44\n" ); +} + +static inline void diag9c (uint64_t val) +{ + asmvolatile ("diag %[val],0,0x9c\n" + : + : [val]"d" (val) + ); +} +#endif
lib/s390x/sie.h +1 -0
@@ -281,6 +281,7 @@ 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 file +26 -0
@@ -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
lib/s390x/snippet-host.h renamed from lib/s390x/snippet.h +6 -3
@@ -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 +144,7 @@ static inline void snippet_setup_guest(struct vm *vm, bool is_pv) } } +bool snippet_check_force_exit (struct vm *vm); +bool snippet_get_force_exit_value (struct vm *vm,uint64_t *value); +void snippet_check_force_exit_value (struct vm *vm,uint64_t exit_exp);#endif
lib/s390x/sie.c +28 -0
@@ -42,6 +42,34 @@ 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) +{ +uint32_t ipb = vm->sblk->ipb; +uint64_t code; +uint16_t displace; +uint8_t base; +bool ret = true; + + ret = ret && vm->sblk->icptcode ==ICPT_INST ; + ret = ret && (vm->sblk->ipa &0xff00 ) ==0x8300 ; +switch (diag) { +case 0x44 : +case 0x9c : + ret = ret && !(ipb &0xffff ); + ipb >>=16 ; + displace = ipb &0xfff ; + ipb >>=12 ; + base = ipb &0xf ; + code = base ? vm->save_area.guest.grs[base] + displace : displace; + code &=0xffff ; + ret = ret && (code == diag); +break ; +default : +abort ();/* not implemented */ + } +return ret; +} +void sie_handle_validity (struct vm *vm) {if (vm->sblk->icptcode !=ICPT_VALIDITY )
lib/s390x/snippet-host.c new file +40 -0
@@ -0,0 +1,40 @@ +/* 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_check_force_exit (struct vm *vm) +{ +bool r; + + r =sie_is_diag_icpt (vm,0x44 ); +report (r,"guest forced exit" ); +return r; +} + +bool snippet_get_force_exit_value (struct vm *vm,uint64_t *value) +{ +struct kvm_s390_sie_block *sblk = vm->sblk; + +if (sie_is_diag_icpt (vm,0x9c )) { + *value = vm->save_area.guest.grs[(sblk->ipa &0xf0 ) >>4 ]; +report_pass ("guest forced exit with value: 0x%lx" , *value); +return true; + } +report_fail ("guest forced exit with value" ); +return false; +} + +void snippet_check_force_exit_value (struct vm *vm,uint64_t value_exp) +{ +uint64_t value; + +if (snippet_get_force_exit_value (vm, &value)) +report (value == value_exp,"guest exit value matches 0x%lx" , value_exp); +}
lib/s390x/uv.c +1 -1
@@ -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
@@ -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
@@ -8,7 +8,7 @@* Janosch Frank <[email protected]> */ #include <libcflat.h> -#include <snippet.h> +#include <snippet-host.h> #include <pv_icptdata.h> #include <sie.h> #include <sclp.h>
s390x/pv-icptcode.c +1 -1
@@ -12,7 +12,7 @@#include <sie.h> #include <smp.h> #include <sclp.h> -#include <snippet.h> +#include <snippet-host.h> #include <pv_icptdata.h> #include <asm/facility.h> #include <asm/barrier.h>
s390x/pv-ipl.c +1 -1
@@ -10,7 +10,7 @@#include <libcflat.h> #include <sie.h> #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 +1 -1
@@ -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
@@ -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
@@ -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
2.41.0
Re: [kvm-unit-tests PATCH 3/5] s390x: Add library functions for exiting from snippet
Quoted 6 lines · On Wed, 13 Dec 2023 13:49:40 +0100 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 diag0x44and0x9cfor this.
Add a guest specific snippet header file and rename the host's.
you should also mention here that you are splitting snippet.h into a host-only part and a guest-only part
Quoted 22 lines
Signed-off-by Nina Schoetterl-Glausch <[email protected]>--- s390x/Makefile | 1 + lib/s390x/asm/arch_def.h | 13 ++++++++ lib/s390x/sie.h | 1 + lib/s390x/snippet-guest.h | 26 ++++++++++++++++ lib/s390x/{snippet.h => snippet-host.h} | 9 ++++-- lib/s390x/sie.c | 28 +++++++++++++++++ lib/s390x/snippet-host.c | 40 +++++++++++++++++++++++++ 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 +- 15 files changed, 123 insertions(+), 11 deletions(-) create mode 100644 lib/s390x/snippet-guest.h rename lib/s390x/{snippet.h => snippet-host.h} (93%) create mode 100644 lib/s390x/snippet-host.c
[...]
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 48
Quoted 12 lines
lib/s390x/sie.c +4
-0@@ -42,6 +42,34 @@ 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) +{ +uint32_t ipb = vm->sblk->ipb; +uint64_t code;
uint64_t code = 0;
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 51
Quoted 3 lines
lib/s390x/sie.c +3
-0+uint16_t displace; +uint8_t base; +bool ret = true;
bool ret;
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 54
Quoted 3 lines
lib/s390x/sie.c +3
-0+ + ret = ret && vm->sblk->icptcode ==ICPT_INST ; + ret = ret && (vm->sblk->ipa &0xff00 ) ==0x8300 ;
ret = vm->sblk->icptcode == ICPT_INST && (vm->sblk->ipa & 0xff00) == 0x8300;
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 60
Quoted 6 lines
lib/s390x/sie.c +6
-0+switch (diag) { +case 0x44 : +case 0x9c : + ret = ret && !(ipb &0xffff ); + ipb >>=16 ; + displace = ipb &0xfff ;
maybe it's more readable to avoid shifting thigs around all the time:
displace = (ipb >> 16) & 0xfff;
base = (ipb >> 28) & 0xf;
if (base)
code = vm->....[base]; code = (code + displace) & 0xffff; if (ipb & 0xffff || code != diag) return false;
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 70
Quoted 10 lines
lib/s390x/sie.c +10
-0+ ipb >>=12 ; + base = ipb &0xf ; + code = base ? vm->save_area.guest.grs[base] + displace : displace; + code &=0xffff ; + ret = ret && (code == diag); +break ; +default : +abort ();/* not implemented */ + } +return ret;
although I have the feeling that this would be more readable if you would check diag immediately, and avoid using ret
Quoted 41 lines
lib/s390x/sie.c +2
-0+} +void sie_handle_validity (struct vm *vm) {if (vm->sblk->icptcode !=ICPT_VALIDITY )lib/s390x/snippet-host.c new file +30
-0@@ -0,0 +1,40 @@ +/* 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_check_force_exit (struct vm *vm) +{ +bool r; + + r =sie_is_diag_icpt (vm,0x44 ); +report (r,"guest forced exit" ); +return r; +} + +bool snippet_get_force_exit_value (struct vm *vm,uint64_t *value) +{ +struct kvm_s390_sie_block *sblk = vm->sblk; + +if (sie_is_diag_icpt (vm,0x9c )) { + *value = vm->save_area.guest.grs[(sblk->ipa &0xf0 ) >>4 ]; +report_pass ("guest forced exit with value: 0x%lx" , *value); +return true; + } +report_fail ("guest forced exit with value" );
lib/s390x/snippet-host.c line 40
Quoted 10 lines
lib/s390x/snippet-host.c new file +10
-0+return false; +} + +void snippet_check_force_exit_value (struct vm *vm,uint64_t value_exp) +{ +uint64_t value; + +if (snippet_get_force_exit_value (vm, &value)) +report (value == value_exp,"guest exit value matches 0x%lx" , value_exp); +}
from a readability and a consistency perspective, it would be better if the functions would only check stuff and return a bool or a value, and do the report() in the body of the testcase
[...]
Re: [kvm-unit-tests PATCH 3/5] s390x: Add library functions for exiting from snippet
Quoted 10 lines · Claudio Imbrenda
Quoted 6 lines · On Wed, 13 Dec 2023 13:49:40 +0100 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 diag0x44and0x9cfor this.
Add a guest specific snippet header file and rename the host's.
Quoted 2 lines
you should also mention here that you are splitting
snippet.hinto a host-only part and a guest-only part
Well, I'm not splitting anything. Is it not clear that "the host's" refers to snippet.h?
How about:
Add a guest specific snippet header file and rename snippet.h to reflect that it is host specific.
Quoted 47 lines
Quoted 23 lines
Signed-off-by Nina Schoetterl-Glausch <[email protected]>--- s390x/Makefile | 1 + lib/s390x/asm/arch_def.h | 13 ++++++++ lib/s390x/sie.h | 1 + lib/s390x/snippet-guest.h | 26 ++++++++++++++++ lib/s390x/{snippet.h => snippet-host.h} | 9 ++++-- lib/s390x/sie.c | 28 +++++++++++++++++ lib/s390x/snippet-host.c | 40 +++++++++++++++++++++++++ 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 +- 15 files changed, 123 insertions(+), 11 deletions(-) create mode 100644 lib/s390x/snippet-guest.h rename lib/s390x/{snippet.h => snippet-host.h} (93%) create mode 100644 lib/s390x/snippet-host.c[...]
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 48
Quoted 12 lines
lib/s390x/sie.c +4
-0@@ -42,6 +42,34 @@ 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) +{ +uint32_t ipb = vm->sblk->ipb; +uint64_t code;
uint64_tcode = 0;lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 51
Quoted 3 lines
lib/s390x/sie.c +3
-0+uint16_t displace; +uint8_t base; +bool ret = true;bool ret;
Quoted 6 lines
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 54
Quoted 3 lines
lib/s390x/sie.c +3
-0+ + ret = ret && vm->sblk->icptcode ==ICPT_INST ; + ret = ret && (vm->sblk->ipa &0xff00 ) ==0x8300 ;ret =
vm->sblk->icptcode==ICPT_INST&& (vm->sblk->ipa&0xff00) ==0x8300;
(*) see below
Quoted 9 lines
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 60
Quoted 7 lines
lib/s390x/sie.c +6
-0+switch (diag) { +case 0x44 : +case 0x9c : + ret = ret && !(ipb &0xffff ); + ipb >>=16 ; + displace = ipb &0xfff ;maybe it's more readable to avoid shifting thigs around all the time:
I don't know, now I gotta be able to do rudimentary arithmetic :D I don't really have a preference.
I wonder if defining a bit field would be worth it.
Quoted 9 lines
displace = (ipb >> 16) &
0xfff;
base = (ipb >> 28) &0xf;
if (base)code = vm->....[base]; code = (code + displace) & 0xffff; if (ipb & 0xffff || code != diag) return false;
Quoted 13 lines
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 70
Quoted 10 lines
lib/s390x/sie.c +10
-0+ ipb >>=12 ; + base = ipb &0xf ; + code = base ? vm->save_area.guest.grs[base] + displace : displace; + code &=0xffff ; + ret = ret && (code == diag); +break ; +default : +abort ();/* not implemented */ + } +return ret;although I have the feeling that this would be more readable if you would check diag immediately, and avoid using ret
Not sure what you mean, do you want an early return at (*)?
Quoted 48 lines
Quoted 48 lines
lib/s390x/sie.c +2
-0+} +void sie_handle_validity (struct vm *vm) {if (vm->sblk->icptcode !=ICPT_VALIDITY )lib/s390x/snippet-host.c new file +36
-0@@ -0,0 +1,40 @@ +/* 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_check_force_exit (struct vm *vm) +{ +bool r; + + r =sie_is_diag_icpt (vm,0x44 ); +report (r,"guest forced exit" ); +return r; +} + +bool snippet_get_force_exit_value (struct vm *vm,uint64_t *value) +{ +struct kvm_s390_sie_block *sblk = vm->sblk; + +if (sie_is_diag_icpt (vm,0x9c )) { + *value = vm->save_area.guest.grs[(sblk->ipa &0xf0 ) >>4 ]; +report_pass ("guest forced exit with value: 0x%lx" , *value); +return true; + } +report_fail ("guest forced exit with value" ); +return false; +} + +void snippet_check_force_exit_value (struct vm *vm,uint64_t value_exp) +{ +uint64_t value;
Quoted 8 lines
lib/s390x/snippet-host.c line 40
Quoted 4 lines
lib/s390x/snippet-host.c new file +4
-0+ +if (snippet_get_force_exit_value (vm, &value)) +report (value == value_exp,"guest exit value matches 0x%lx" , value_exp); +}from a readability and a consistency perspective, it would be better if the functions would only check stuff and return a bool or a value, and do the
report()in the body of the testcase
Hmm, I chose to do the report in order to be consistent with check_pgm_int_code.
Quoted 3 lines
[...]
Re: [kvm-unit-tests PATCH 3/5] s390x: Add library functions for exiting from snippet
Quoted 14 lines · On Thu, 14 Dec 2023 21:02:53 +0100 Nina Schoetterl-Glausch <[email protected]>…
Quoted 10 lines · Claudio Imbrenda
Quoted 6 lines · On Wed, 13 Dec 2023 13:49:40 +0100 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 diag0x44and0x9cfor this.
Add a guest specific snippet header file and rename the host's.Quoted 2 lines
you should also mention here that you are splitting
snippet.hinto a host-only part and a guest-only part
Quoted 6 lines
Well, I'm not splitting anything. Is it not clear that "the host's" refers to
snippet.h?How about:
Add a guest specific snippet header file and renamesnippet.hto reflect that it is host specific.
sounds good
[...]
Quoted 21 lines
Quoted 21 lines
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 48
Quoted 12 lines
lib/s390x/sie.c +4
-0@@ -42,6 +42,34 @@ 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) +{ +uint32_t ipb = vm->sblk->ipb; +uint64_t code;
uint64_tcode = 0;lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 51
Quoted 3 lines
lib/s390x/sie.c +3
-0+uint16_t displace; +uint8_t base; +bool ret = true;bool ret;
Quoted 4 lines
Quoted 4 lines
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 54
Quoted 3 lines
lib/s390x/sie.c +3
-0+ + ret = ret && vm->sblk->icptcode ==ICPT_INST ; + ret = ret && (vm->sblk->ipa &0xff00 ) ==0x8300 ;
something like this:
assert(diag == 0x44 || diag == 0x9c);
if (vm->sblk->icptcode != ICPT_INST)
return false; if ((vm->sblk->ipa & 0xff00) != 0x8300) return false; if (vm->sblk->ipb & 0xffff) return false;
code = ....
return code == diag;
Quoted 14 lines
Quoted 2 lines
ret =
vm->sblk->icptcode==ICPT_INST&& (vm->sblk->ipa&0xff00) ==0x8300;(*) see below
Quoted 9 lines
lib/s390x/sie.c void sie_check_validity(struct vm *vm, uint16_t vir_exp) · line 60
Quoted 7 lines
lib/s390x/sie.c +6
-0+switch (diag) { +case 0x44 : +case 0x9c : + ret = ret && !(ipb &0xffff ); + ipb >>=16 ; + displace = ipb &0xfff ;maybe it's more readable to avoid shifting thigs around all the time:
Quoted 3 lines
I don't know, now I gotta be able to do rudimentary arithmetic :D I don't really have a preference.
I wonder if defining a bit field would be worth it.
I think it would.
maybe something like:
union ip_text {
struct {
unsigned long ipa:16;
unsigned long ipb:32;
};
struct {
unsigned long opcode:8;
...
};
}then you can do this at the beginning of the function:
union ip_text ip = { .ipa = vm->sblk->ipa, .ipb = ... };
and then use only the bitfields
[...]
PATCH 4/5 s390x: Use library functions for snippet exit
In reply to Nina Schoetterl-GlauschReplace the existing code for exiting from snippets with the newly introduced library functionality.
This causes additional report()s, otherwise no change in functionality intended.
--- s390x/sie-dat.c | 10 ++-------- s390x/snippets/c/sie-dat.c | 19 +------------------ 2 files changed, 3 insertions(+), 26 deletions(-)
s390x/sie-dat.c +2 -8
@@ -27,23 +27,17 @@ 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_get_force_exit_value (&vm, &test_page_gpa)); 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_check_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
@@ -9,28 +9,11 @@*/ #include <libcflat.h> #include <asm-generic/page.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 ) -{ - asmvolatile ("diag 0,0,0x44\n" - : - : - :"memory" - ); -} - -static inline void force_exit_value (uint64_t val) -{ - asmvolatile ("diag %[val],0,0x9c\n" - : - : [val]"d" (val) - :"memory" - ); -} -int main (void ) {uint8_t *invalid_ptr;
Signature · 1 line
2.41.0
Re: [kvm-unit-tests PATCH 4/5] s390x: Use library functions for snippet exit
Quoted 19 lines · On Wed, 13 Dec 2023 13:49:41 +0100 Nina Schoetterl-Glausch <[email protected]>…
Replace the existing code for exiting from snippets with the newly introduced library functionality.
This causes additional report()s, otherwise no change in functionality intended.Signed-off-by Nina Schoetterl-Glausch <[email protected]>--- s390x/sie-dat.c | 10 ++-------- s390x/snippets/c/sie-dat.c | 19 +------------------ 2 files changed, 3 insertions(+), 26 deletions(-)s390x/sie-dat.c +0
-0@@ -27,23 +27,17 @@ static void test_sie_dat(void)uint64_t test_page_gpa, test_page_hpa;uint8_t *test_page_hva, expected_val;bool contents_match;
s390x/sie-dat.c static void test_sie_dat(void) · line 33
Quoted 10 lines
s390x/sie-dat.c +1
-6-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_get_force_exit_value (&vm, &test_page_gpa));
but the function inside the assert will already report a failure, right? then why the assert? (the point I'm trying to make is that the function should not report stuff, see my comments in the previous patch)
Quoted 46 lines
s390x/sie-dat.c +1
-2test_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_check_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@@ -9,28 +9,11 @@*/ #include <libcflat.h> #include <asm-generic/page.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 ) -{ - asmvolatile ("diag 0,0,0x44\n" - : - : - :"memory" - ); -} - -static inline void force_exit_value (uint64_t val) -{ - asmvolatile ("diag %[val],0,0x9c\n" - : - : [val]"d" (val) - :"memory" - ); -} -int main (void ) {uint8_t *invalid_ptr;
Re: [kvm-unit-tests PATCH 4/5] s390x: Use library functions for snippet exit
Quoted 24 lines · Claudio Imbrenda
Quoted 21 lines · On Wed, 13 Dec 2023 13:49:41 +0100 Nina Schoetterl-Glausch <[email protected]>…
Replace the existing code for exiting from snippets with the newly introduced library functionality.
This causes additional report()s, otherwise no change in functionality intended.Signed-off-by Nina Schoetterl-Glausch <[email protected]>--- s390x/sie-dat.c | 10 ++-------- s390x/snippets/c/sie-dat.c | 19 +------------------ 2 files changed, 3 insertions(+), 26 deletions(-)s390x/sie-dat.c +0
-1@@ -27,23 +27,17 @@ 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;
Quoted 11 lines
s390x/sie-dat.c static void test_sie_dat(void) · line 33
Quoted 8 lines
s390x/sie-dat.c +1
-5/* 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_get_force_exit_value (&vm, &test_page_gpa));but the function inside the assert will already report a failure, right? then why the assert? (the point I'm trying to make is that the function
The assert makes sense, if it fails something has gone
completely off the rails. The question is indeed rather if to report. There is no harm in it, but I'm thinking now that there should be _get_ functions that don't report and optionally also _check_ functions that do.
So:snippet_get_force_exitsnippet_check_force_exit (exists, but only the _get_ variant is currently required) snippet_get_force_exit_value (exists, required) snippet_check_force_exit_value (exists, but unused)
So minimally, we could do:snippet_get_force_exitsnippet_get_force_exit_value
I'm thinking we should go for the following:
bool snippet_has_force_exit(...)
bool snippet_has_force_exit_value(...)uint64_t snippet_get_force_exit_value(...) (internally does assert(snippet_has_forced_exit_value)) void snippet_check_force_exit_value(...) (or whoever needs this in the future adds it) (Naming open to suggestions)
Then this becomes:
/* 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);
-
+ assert(snippet_has_force_exit_value(&vm);
- r1 = (vm.sblk->ipa & 0xf0) >> 4;
- test_page_gpa = vm.save_area.guest.grs[r1];
+ test_page_gpa = snippet_get_force_exit_value(&vm);
Quoted 14 lines
should not report stuff, see my comments in the previous patch)
Quoted 12 lines
s390x/sie-dat.c +1
-2test_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_check_force_exit (&vm)); contents_match = true;for (unsigned int i =0 ; i <GUEST_TEST_PAGE_COUNT ; i++) {
[...]
Re: [kvm-unit-tests PATCH 4/5] s390x: Use library functions for snippet exit
Quoted 52 lines · On Fri, 15 Dec 2023 12:50:15 +0100 Nina Schoetterl-Glausch <[email protected]>…
Quoted 24 lines · Claudio Imbrenda
Quoted 21 lines · On Wed, 13 Dec 2023 13:49:41 +0100 Nina Schoetterl-Glausch <[email protected]>…
Replace the existing code for exiting from snippets with the newly introduced library functionality.
This causes additional report()s, otherwise no change in functionality intended.Signed-off-by Nina Schoetterl-Glausch <[email protected]>--- s390x/sie-dat.c | 10 ++-------- s390x/snippets/c/sie-dat.c | 19 +------------------ 2 files changed, 3 insertions(+), 26 deletions(-)s390x/sie-dat.c +0
-1@@ -27,23 +27,17 @@ 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;Quoted 11 lines
s390x/sie-dat.c static void test_sie_dat(void) · line 33
Quoted 8 lines
s390x/sie-dat.c +1
-5/* 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_get_force_exit_value (&vm, &test_page_gpa));but the function inside the assert will already report a failure, right? then why the assert? (the point I'm trying to make is that the function
The assert makes sense, if it fails something has gone
completely off the rails. The question is indeed rather if to report. There is no harm in it, but I'm thinking now that there should be _get_ functions that don't report and optionally also _check_ functions that do.So:
snippet_get_force_exitsnippet_check_force_exit(exists, but only the _get_ variant is currently required)snippet_get_force_exit_value(exists, required)snippet_check_force_exit_value(exists, but unused)So minimally, we could do:
snippet_get_force_exitsnippet_get_force_exit_value
Quoted 6 lines
I'm thinking we should go for the following:
boolsnippet_has_force_exit(...)
boolsnippet_has_force_exit_value(...)uint64_tsnippet_get_force_exit_value(...)(internally doesassert(snippet_has_forced_exit_value)) voidsnippet_check_force_exit_value(...)(or whoever needs this in the future adds it) (Naming open to suggestions)
I like this.
I would call the bool ones "snippet_is_force_exit", in line with similar functions. usually "has" is used for capabilities or features.
but I'm open to other suggestions/ideas
Quoted 28 lines
Then this becomes:
s390x/sie-dat.c +2
-5/* 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 ); - + assert(snippet_has_force_exit_value(&vm); - r1 = (vm.sblk->ipa &0xf0 ) >>4 ; - test_page_gpa = vm.save_area.guest.grs[r1]; + test_page_gpa =snippet_get_force_exit_value (&vm);Quoted 15 lines
should not report stuff, see my comments in the previous patch)
Quoted 12 lines
s390x/sie-dat.c +1
-2test_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_check_force_exit (&vm)); contents_match = true;for (unsigned int i =0 ; i <GUEST_TEST_PAGE_COUNT ; i++) {[...]
PATCH 5/5 s390x: Add test for STFLE interpretive execution (format-0)
In reply to Nina Schoetterl-GlauschThe 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 | 132 +++++++++++++++++++++++++++++++++++++++ 4 files changed, 169 insertions(+), 1 deletion(-) create mode 100644 s390x/snippets/c/stfle.c create mode 100644 s390x/stfle-sie.c
s390x/Makefile +2 -0
@@ -42,6 +42,7 @@ 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
@@ -127,6 +128,7 @@ 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
@@ -27,12 +27,20 @@ static inline void stfl(void) asmvolatile (" 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; asmvolatile (" .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
@@ -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 ; + asmvolatile ("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 +132 -0
132 lines not on this page · raw message
Signature · 1 line
2.41.0
Re: [kvm-unit-tests PATCH 5/5] s390x: Add test for STFLE interpretive execution (format-0)
Quoted 6 lines · On Wed, 13 Dec 2023 13:49:42 +0100 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.Signed-off-by Nina Schoetterl-Glausch <[email protected]>
[...]
Quoted 9 lines
static inline void
setup_facilities(void)s390x/snippets/c/stfle.c new file +2
-0@@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/*
s390x/snippets/c/stfle.c line 12
Quoted 10 lines
s390x/snippets/c/stfle.c new file +10
-0+* 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 ;
why 8?
Quoted 68 lines
s390x/snippets/c/stfle.c new file +14
-0+uint64_t res[max_fac_len +1 ]; + + res[0 ] = max_fac_len -1 ; + asmvolatile ("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 +48
-0@@ -0,0 +1,132 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* Copyright IBM Corp. 2023 +* +* SIE with STLFE interpretive execution facilities test. +*/ +#include <libcflat.h> +#include <stdlib.h> +#include <asm/facility.h> +#include <asm/time.h> +#include <snippet-host.h> +#include <alloc_page.h> +#include <sclp.h> + +static struct vm vm; +static uint64_t (*fac)[PAGE_SIZE /sizeof (uint64_t )]; +static rand_state rand_s; + +static void setup_guest (void ) +{ +extern const char SNIPPET_NAME_START (c ,stfle )[]; +extern const char SNIPPET_NAME_END (c ,stfle )[]; + +setup_vm (); + fac =alloc_pages_flags (0 ,AREA_DMA31 ); + +snippet_setup_guest (&vm, false); +snippet_init (&vm,SNIPPET_NAME_START (c, stfle), +SNIPPET_LEN (c, stfle),SNIPPET_UNPACK_OFF ); +} + +struct guest_stfle_res { +uint16_t len; +uint64_t reg; +unsigned char *mem; +}; + +static struct guest_stfle_res run_guest (void ) +{ +struct guest_stfle_res res; +uint64_t guest_stfle_addr; + +sie (&vm); +assert (snippet_get_force_exit_value (&vm, &guest_stfle_addr)); + res.mem = &vm.guest_mem[guest_stfle_addr]; +memcpy (&res.reg, res.mem,sizeof (res.reg)); + res.len = (res.reg &0xff ) +1 ; + res.mem +=sizeof (res.reg);
s390x/stfle-sie.c line 58
Quoted 10 lines
s390x/stfle-sie.c new file +10
-0+return res; +} + +static void test_stfle_format_0 (void ) +{ +struct guest_stfle_res res; + +report_prefix_push ("format-0" ); +for (int j =0 ; j <stfle_size (); j++) +WRITE_ONCE ((*fac)[j],rand64 (&rand_s));
do you really need random numbers? can't you use a static pattern? maybe something like 0x0001020304050607 etc...
s390x/stfle-sie.c line 63
Quoted 5 lines
s390x/stfle-sie.c new file +5
-0+ vm.sblk->fac = (uint32_t )(uint64_t )fac; + res =run_guest (); +report (res.len ==stfle_size (),"stfle len correct" ); +report (!memcmp (*fac, res.mem, res.len *sizeof (uint64_t )), +"Guest facility list as specified" );
it seems like you are comparing the full facility list (stfle_size doublewords long) with the result of STFLE in the guest, but the guest is limited to 8 double words?
Quoted 41 lines
s390x/stfle-sie.c new file +41
-0+report_prefix_pop (); +} + +struct args { +uint64_t seed; +}; + +static bool parse_uint64_t (const char *arg,uint64_t *out) +{ +char *end; +uint64_t num; + +if (arg[0 ] =='\0' ) +return false; + num =strtoul (arg, &end,0 ); +if (end[0 ] !='\0' ) +return false; + *out = num; +return true; +} + +static struct args parse_args (int argc,char **argv) +{ +struct args args; +const char *flag; +unsigned int i; +uint64_t arg; +bool has_arg; + +stck (&args.seed); + +for (i =1 ; i < argc; i++) { +if (i +1 < argc) + has_arg =parse_uint64_t (argv[i +1 ], &arg); +else + has_arg = false; + + flag ="--seed" ; +if (!strcmp (flag, argv[i])) { +if (!has_arg) +report_abort ("%s needs an uint64_t parameter" , flag);
s390x/stfle-sie.c line 114
Quoted 10 lines
s390x/stfle-sie.c new file +10
-0+ args.seed = arg; + ++i; +continue ; + } +report_abort ("Unsupported parameter '%s'" , + argv[i]); + } + +return args; +}
this is lots of repeated code in all tests, I should really resurrect and polish my patchseries for argument parsing
Quoted 18 lines
s390x/stfle-sie.c new file +18
-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; + } + +report_info ("pseudo rand seed: 0x%lx" , args.seed); + rand_s =RAND_STATE_INIT (args.seed); +setup_guest (); +if (test_facility (7 )) +test_stfle_format_0 (); +out: +return report_summary (); +}
Re: [kvm-unit-tests PATCH 5/5] s390x: Add test for STFLE interpretive execution (format-0)
Quoted 25 lines · Claudio Imbrenda
Quoted 6 lines · On Wed, 13 Dec 2023 13:49:42 +0100 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.Signed-off-by Nina Schoetterl-Glausch <[email protected]>[...]
Quoted 13 lines
static inline void
setup_facilities(void)s390x/snippets/c/stfle.c new file +6
-0@@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* Copyright IBM Corp. 2023 +* +* Snippet used by the STLFE interpretive execution facilities test. +*/
Quoted 8 lines
s390x/snippets/c/stfle.c line 12
Quoted 6 lines
s390x/snippets/c/stfle.c new file +6
-0+#include <libcflat.h> +#include <snippet-guest.h> + +int main (void ) +{ +const unsigned int max_fac_len =8 ;why 8?
8 is a somewhat arbitrary, large number :)
I suppose I could choose an even larger one, maybe even PAGE_SIZE/8. That would guarantee that max_fac_len >= stfle_size() (8 is enough for that today) It's not necessary for max_fac_len >= stfle_size(), but probably good for test coverage.
Quoted 21 lines
Quoted 21 lines
s390x/snippets/c/stfle.c new file +14
-0+uint64_t res[max_fac_len +1 ]; + + res[0 ] = max_fac_len -1 ; + asmvolatile ("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 +0
-0@@ -0,0 +1,132 @@
[...]
Quoted 11 lines
s390x/stfle-sie.c line 58
Quoted 8 lines
s390x/stfle-sie.c new file +8
-0+ +static void test_stfle_format_0 (void ) +{ +struct guest_stfle_res res; + +report_prefix_push ("format-0" ); +for (int j =0 ; j <stfle_size (); j++) +WRITE_ONCE ((*fac)[j],rand64 (&rand_s));do you really need random numbers? can't you use a static pattern? maybe something like
0x0001020304050607etc...
Doesn't really need to be random, I need some arbitrary test pattern, but I don't think some cumbersome constant literal improves anything. The RNG is just initialized with the time, because why not.
Quoted 4 lines
s390x/stfle-sie.c line 61
Quoted 4 lines
s390x/stfle-sie.c new file +3
-0+ vm.sblk->fac = (uint32_t )(uint64_t )fac; + res =run_guest (); +report (res.len ==stfle_size (),"stfle len correct" );
^ should be
+ report(res.len == min(stfle_size(), 8), "stfle len correct");
For the case that the guest buffer was shorter.
Quoted 6 lines
s390x/stfle-sie.c line 63
Quoted 2 lines
s390x/stfle-sie.c new file +2
-0+report (!memcmp (*fac, res.mem, res.len *sizeof (uint64_t )), +"Guest facility list as specified" );it seems like you are comparing the full facility list (
stfle_sizedoublewords long) with the result of STFLE in the guest, but the guest is limited to 8 double words?
Their prefixes must be the same. res.len is the guest length, so max 8 right now.
Quoted 3 lines
Quoted 3 lines
s390x/stfle-sie.c new file +2
-0+report_prefix_pop (); +}
Re: [kvm-unit-tests PATCH 5/5] s390x: Add test for STFLE interpretive execution (format-0)
Quoted 61 lines · Nina Schoetterl-Glausch
Quoted 25 lines · Claudio Imbrenda
Quoted 6 lines · On Wed, 13 Dec 2023 13:49:42 +0100 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.Signed-off-by Nina Schoetterl-Glausch <[email protected]>[...]
Quoted 13 lines
static inline void
setup_facilities(void)s390x/snippets/c/stfle.c new file +6
-0@@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* +* Copyright IBM Corp. 2023 +* +* Snippet used by the STLFE interpretive execution facilities test. +*/ Quoted 8 lines
s390x/snippets/c/stfle.c line 12
Quoted 6 lines
s390x/snippets/c/stfle.c new file +6
-0+#include <libcflat.h> +#include <snippet-guest.h> + +int main (void ) +{ +const unsigned int max_fac_len =8 ;why 8?
8 is a somewhat arbitrary, large number :)
I suppose I could choose an even larger one, maybe even PAGE_SIZE/8. That would guarantee thatmax_fac_len>=stfle_size()(8 is enough for that today) It's not necessary formax_fac_len>=stfle_size(), but probably good for test coverage.Quoted 21 lines
Quoted 21 lines
s390x/snippets/c/stfle.c new file +14
-0+uint64_t res[max_fac_len +1 ]; + + res[0 ] = max_fac_len -1 ; + asmvolatile ("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 +0
-0@@ -0,0 +1,132 @@
[...]
Quoted 3 lines
Quoted 3 lines
s390x/stfle-sie.c line 61
Quoted 3 lines
s390x/stfle-sie.c new file +3
-0+ vm.sblk->fac = (uint32_t )(uint64_t )fac; + res =run_guest (); +report (res.len ==stfle_size (),"stfle len correct" );
You're right, disregard everything below.
Quoted 18 lines
^ should be
(quoted diff) +1
-0+ report(res.len == min(stfle_size(), 8), "stfle len correct");For the case that the guest buffer was shorter.
Quoted 6 lines
s390x/stfle-sie.c line 63
Quoted 2 lines
s390x/stfle-sie.c new file +2
-0+report (!memcmp (*fac, res.mem, res.len *sizeof (uint64_t )), +"Guest facility list as specified" );it seems like you are comparing the full facility list (
stfle_sizedoublewords long) with the result of STFLE in the guest, but the guest is limited to 8 double words?Their prefixes must be the same. res.len is the guest length, so max 8 right now.
Quoted 3 lines
Quoted 3 lines
s390x/stfle-sie.c new file +2
-0+report_prefix_pop (); +}