s390x: Add snippet support
cover s390x: Add snippet support
The SIE support allows us to run guests and test the hypervisor's (V)SIE implementation. However it requires that the guest instructions are binary which limits the complexity of the guest code.
The snippet support provides a way to write guest code as ASM or C and simply memcpy it into guest memory. Some of the KVM-unit-test library can be re-used which further speeds up guest code development.
The included mvpg-sie test helped us to deliver the KVM mvpg fixes which Claudio posted a short while ago. In the future I'll post Secure Execution snippet support patches which was my initial goal with this series anyway.
RFC -> V1:
* Replaced handle_validity() with assert()
* Added clearing of registers before sigp in cstart.S
* Added a nice way to compile the snippets
Janosch Frank (2):
s390x: snippets: Add gitignore as well as linker script and start
assembly
s390x: mvpg: Add SIE mvpg test
Steffen Eiden (1):
s390x: snippets: Add snippet compilation
.gitignore | 1 + s390x/Makefile | 29 ++++-- s390x/mvpg-sie.c | 150 ++++++++++++++++++++++++++++++++ s390x/snippets/c/cstart.S | 15 ++++ s390x/snippets/c/flat.lds | 51 +++++++++++ s390x/snippets/c/mvpg-snippet.c | 33 +++++++ s390x/unittests.cfg | 3 + 7 files changed, 276 insertions(+), 6 deletions(-) create mode 100644 s390x/mvpg-sie.c create mode 100644 s390x/snippets/c/cstart.S create mode 100644 s390x/snippets/c/flat.lds create mode 100644 s390x/snippets/c/mvpg-snippet.c
Signature · 1 line
2.27.0
PATCH 1/3 s390x: snippets: Add gitignore as well as linker script and start assembly
Snippets are small guests That can be run under a unit test as the hypervisor. They can be written in C or assembly. The C code needs a linker script and a start assembly file that jumps to main to work properly. So let's add that as well as a gitignore entry for the new files.
--- .gitignore | 1 + s390x/snippets/c/cstart.S | 15 ++++++++++++ s390x/snippets/c/flat.lds | 51 +++++++++++++++++++++++++++++++++++++++ 3 files changed, 67 insertions(+) create mode 100644 s390x/snippets/c/cstart.S create mode 100644 s390x/snippets/c/flat.lds
.gitignore +1 -0
@@ -23,3 +23,4 @@ cscope.* /api/dirty-log /api/dirty-log-perf /s390x/*.bin +/s390x/snippets/*/*.gbin
s390x/snippets/c/cstart.S new file +15 -0
@@ -0,0 +1,15 @@ +#include <asm/sigp.h> + +.section .init + .globl start +start: + /* XOR all registers with themselves to clear them fully. */ + .irp i, 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 + xgr \i,\i + .endr + /* 0x3000 is the stack page for now */ + lghi %r15, 0x4000 - 160 + brasl %r14, main + /* For now let's only use cpu 0 in snippets so this will always work. */ + xgr %r0, %r0 + sigp %r2, %r0, SIGP_STOP
s390x/snippets/c/flat.lds new file +51 -0
51 lines not on this page · raw message
Signature · 1 line
2.27.0
Re: [PATCH 1/3] s390x: snippets: Add gitignore as well as linker script and start assembly
Quoted 51 lines · Janosch Frank
Snippets are small guests That can be run under a unit test as the hypervisor. They can be written in C or assembly. The C code needs a linker script and a start assembly file that jumps to main to work properly. So let's add that as well as a gitignore entry for the new files.
Signed-off-by Janosch Frank <[email protected]>--- .gitignore | 1 + s390x/snippets/c/cstart.S | 15 ++++++++++++ s390x/snippets/c/flat.lds | 51 +++++++++++++++++++++++++++++++++++++++ 3 files changed, 67 insertions(+)create mode 100644
s390x/snippets/c/cstart.S
create mode 100644s390x/snippets/c/flat.lds.gitignore +1
-0@@ -23,3 +23,4 @@ cscope.* /api/dirty-log /api/dirty-log-perf /s390x/*.bin +/s390x/snippets/*/*.gbins390x/snippets/c/cstart.S new file +15
-0@@ -0,0 +1,15 @@ +#include <asm/sigp.h> + +.section .init + .globl start +start: + /* XOR all registers with themselves to clear them fully. */ + .irp i, 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 + xgr \i,\i + .endr + /* 0x3000 is the stack page for now */ + lghi %r15, 0x4000 - 160 + brasl %r14, main + /* For now let's only use cpu 0 in snippets so this will always work. */ + xgr %r0, %r0 + sigp %r2, %r0, SIGP_STOPs390x/snippets/c/flat.lds new file +0
-0
s390x/snippets/c/flat.lds line 7
Quoted 7 lines
s390x/snippets/c/flat.lds new file +7
-0@@ -0,0 +1,51 @@ +SECTIONS +{ + .lowcore : { + /* + * Initial short psw for disk boot, with 31 bit addressing for + * non z/Arch environment compatibility and the instruction + * address 0x10000 (cstart64.S .init).
I think this comment needs some adjustments (0x10000 => 0x4000 and do not talk about cstart64.S)?
Also, what about switching to 64-bit mode in the snippets?
Thomas
Quoted 45 lines
s390x/snippets/c/flat.lds new file +44
-0+ */ + . = 0; + LONG(0x00080000) + LONG(0x80004000) + /* Restart new PSW for booting via PSW restart. */ + . = 0x1a0; + QUAD(0x0000000180000000) + QUAD(0x0000000000004000) + } + . = 0x4000; + .text : { + *(.init) + *(.text) + *(.text.*) + } + . = ALIGN(64K); + etext = .; + .opd : { *(.opd) } + . = ALIGN(16); + .dynamic : { + dynamic_start = .; + *(.dynamic) + } + .dynsym : { + dynsym_start = .; + *(.dynsym) + } + .rela.dyn : { *(.rela*) } + . = ALIGN(16); + .data : { + *(.data) + *(.data.rel*) + } + . = ALIGN(16); + .rodata : { *(.rodata) *(.rodata.*) } + . = ALIGN(16); + __bss_start = .; + .bss : { *(.bss) } + __bss_end = .; + . = ALIGN(64K); + edata = .; + . += 64K; + . = ALIGN(64K); +}
Re: [PATCH 1/3] s390x: snippets: Add gitignore as well as linker script and start assembly
Quoted 54 lines · Thomas Huth
Quoted 53 lines · Janosch Frank
Snippets are small guests That can be run under a unit test as the hypervisor. They can be written in C or assembly. The C code needs a linker script and a start assembly file that jumps to main to work properly. So let's add that as well as a gitignore entry for the new files.
Signed-off-by Janosch Frank <[email protected]>--- .gitignore | 1 + s390x/snippets/c/cstart.S | 15 ++++++++++++ s390x/snippets/c/flat.lds | 51 +++++++++++++++++++++++++++++++++++++++ 3 files changed, 67 insertions(+)create mode 100644
s390x/snippets/c/cstart.S
create mode 100644s390x/snippets/c/flat.lds.gitignore +1
-0@@ -23,3 +23,4 @@ cscope.* /api/dirty-log /api/dirty-log-perf /s390x/*.bin +/s390x/snippets/*/*.gbins390x/snippets/c/cstart.S new file +15
-0@@ -0,0 +1,15 @@ +#include <asm/sigp.h> + +.section .init + .globl start +start: + /* XOR all registers with themselves to clear them fully. */ + .irp i, 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15 + xgr \i,\i + .endr + /* 0x3000 is the stack page for now */ + lghi %r15, 0x4000 - 160 + brasl %r14, main + /* For now let's only use cpu 0 in snippets so this will always work. */ + xgr %r0, %r0 + sigp %r2, %r0, SIGP_STOPs390x/snippets/c/flat.lds new file +2
-0@@ -0,0 +1,51 @@ +SECTIONS +{
Quoted 8 lines
s390x/snippets/c/flat.lds line 7
Quoted 5 lines
s390x/snippets/c/flat.lds new file +5
-0+ .lowcore : { + /* + * Initial short psw for disk boot, with 31 bit addressing for + * non z/Arch environment compatibility and the instruction + * address 0x10000 (cstart64.S .init).I think this comment needs some adjustments (
0x10000=>0x4000and do not talk aboutcstart64.S)?
Right, will do
Quoted 2 lines
Also, what about switching to 64-bit mode in the snippets?
I thought about that for some time a while ago.
It's not really necessary since the host test can also decide which guest PSW is used to start the snippet but it also doesn't hurt so I just added a sam64.
Quoted 50 lines
Thomas
Quoted 45 lines
s390x/snippets/c/flat.lds new file +44
-0+ */ + . = 0; + LONG(0x00080000) + LONG(0x80004000) + /* Restart new PSW for booting via PSW restart. */ + . = 0x1a0; + QUAD(0x0000000180000000) + QUAD(0x0000000000004000) + } + . = 0x4000; + .text : { + *(.init) + *(.text) + *(.text.*) + } + . = ALIGN(64K); + etext = .; + .opd : { *(.opd) } + . = ALIGN(16); + .dynamic : { + dynamic_start = .; + *(.dynamic) + } + .dynsym : { + dynsym_start = .; + *(.dynsym) + } + .rela.dyn : { *(.rela*) } + . = ALIGN(16); + .data : { + *(.data) + *(.data.rel*) + } + . = ALIGN(16); + .rodata : { *(.rodata) *(.rodata.*) } + . = ALIGN(16); + __bss_start = .; + .bss : { *(.bss) } + __bss_end = .; + . = ALIGN(64K); + edata = .; + . += 64K; + . = ALIGN(64K); +}
PATCH 2/3 s390x: snippets: Add snippet compilation
In reply to Janosch FrankFrom: Steffen Eiden <[email protected]>
This patch provides the necessary make targets to properly compile snippets and link them into their associated host.
To define the guest-host definition, we use the make-feature SECONDEXPANSION in combination with Target specific Variable Values. The variable snippets has different values depending on the current target. This enables us to define which snippets (=guests) belong to which hosts. Furthermore, using the second-expansion, we can use snippets in the perquisites of the host's %.elf rule, which otherwise would be not set.
--- s390x/Makefile | 27 +++++++++++++++++++++------ 1 file changed, 21 insertions(+), 6 deletions(-)
s390x/Makefile +21 -6
@@ -76,11 +76,26 @@ OBJDIRS += lib/s390x asmlib = $(TEST_DIR)/cstart64.o $(TEST_DIR)/cpu.o FLATLIBS = $(libcflat) -%.elf: %.o $(FLATLIBS) $(SRCDIR)/s390x/flat.lds $(asmlib) - $(CC) $(CFLAGS) -c -o $(@:.elf=.aux.o) \ - $(SRCDIR)/lib/auxinfo.c -DPROGNAME=\"$@\" - $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/flat.lds \ - $(filter %.o, $^) $(FLATLIBS) $(@:.elf=.aux.o) + +SNIPPET_DIR = $(TEST_DIR)/snippets +snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o + +# perquisites (=guests) for the snippet hosts. +# $(TEST_DIR)/<snippet-host>.elf: snippets = $(SNIPPET_DIR)/<c/asm>/<snippet>.gbin + +$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o $(FLATLIBS) + $(OBJCOPY) -O binary $(patsubst %.gbin,%.o,$@) $@ + $(OBJCOPY) -I binary -O elf64-s390 -B "s390:64-bit" $@ $@ + +$(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_asmlib) $(FLATLIBS) + $(CC) $(LDFLAGS) -o $@ -T $(SNIPPET_DIR)/c/flat.lds $(patsubst %.gbin,%.o,$@) $(snippet_asmlib) $(FLATLIBS) + $(OBJCOPY) -O binary $@ $@ + $(OBJCOPY) -I binary -O elf64-s390 -B "s390:64-bit" $@ $@ + +.SECONDEXPANSION: +%.elf: $$(snippets) %.o $(FLATLIBS) $(SRCDIR)/s390x/flat.lds $(asmlib) + $(CC) $(CFLAGS) -c -o $(@:.elf=.aux.o) $(SRCDIR)/lib/auxinfo.c -DPROGNAME=\"$@\" + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/flat.lds $(filter %.o, $^) $(FLATLIBS) $(snippets) $(@:.elf=.aux.o) $(RM) $(@:.elf=.aux.o) @chmod a-x $@
@@ -94,7 +109,7 @@ FLATLIBS = $(libcflat) $(GENPROTIMG) --host-key-document $(HOST_KEY_DOCUMENT) --no-verify --image $< -o $@ arch_clean: asm_offsets_clean - $(RM) $(TEST_DIR)/*.{o,elf,bin} $(TEST_DIR)/.*.d lib/s390x/.*.d + $(RM) $(TEST_DIR)/*.{o,elf,bin} $(TEST_DIR)/.*.d $(SNIPPET_DIR)/c/*.{o,gbin} $(SNIPPET_DIR)/c/.*.d lib/s390x/.*.d generated-files = $(asm-offsets) $(tests:.elf=.o) $(asmlib) $(cflatobjs): $(generated-files)
Signature · 1 line
2.27.0
Re: [PATCH 2/3] s390x: snippets: Add snippet compilation
Quoted 41 lines · Janosch Frank
From: Steffen Eiden <[email protected]>
This patch provides the necessary make targets to properly compile snippets and link them into their associated host.
To define the guest-host definition, we use the make-feature
SECONDEXPANSIONin combination withTarget specific Variable Values. The variablesnippetshas different values depending on the current target. This enables us to define which snippets (=guests) belong to which hosts. Furthermore, using the second-expansion, we can usesnippetsin the perquisites of the host's%.elfrule, which otherwise would be not set.Signed-off-by Steffen Eiden <[email protected]>--- s390x/Makefile | 27 +++++++++++++++++++++------ 1 file changed, 21 insertions(+), 6 deletions(-)s390x/Makefile +10
-5@@ -76,11 +76,26 @@ OBJDIRS += lib/s390x asmlib = $(TEST_DIR)/cstart64.o $(TEST_DIR)/cpu.o FLATLIBS = $(libcflat) -%.elf: %.o $(FLATLIBS) $(SRCDIR)/s390x/flat.lds $(asmlib) - $(CC) $(CFLAGS) -c -o $(@:.elf=.aux.o) \ - $(SRCDIR)/lib/auxinfo.c -DPROGNAME=\"$@\" - $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/flat.lds \ - $(filter %.o, $^) $(FLATLIBS) $(@:.elf=.aux.o) + +SNIPPET_DIR = $(TEST_DIR)/snippets +snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o + +# perquisites (=guests) for the snippet hosts. +# $(TEST_DIR)/<snippet-host>.elf: snippets = $(SNIPPET_DIR)/<c/asm>/<snippet>.gbin + +$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o $(FLATLIBS) + $(OBJCOPY) -O binary $(patsubst %.gbin,%.o,$@) $@ + $(OBJCOPY) -I binary -O elf64-s390 -B "s390:64-bit" $@ $@
s390x/Makefile OBJDIRS += lib/s390x · line 98
Quoted 10 lines
s390x/Makefile +10
-0+ +$(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_asmlib) $(FLATLIBS) + $(CC) $(LDFLAGS) -o $@ -T $(SNIPPET_DIR)/c/flat.lds $(patsubst %.gbin,%.o,$@) $(snippet_asmlib) $(FLATLIBS) + $(OBJCOPY) -O binary $@ $@ + $(OBJCOPY) -I binary -O elf64-s390 -B "s390:64-bit" $@ $@ + +.SECONDEXPANSION: +%.elf: $$(snippets) %.o $(FLATLIBS) $(SRCDIR)/s390x/flat.lds $(asmlib) + $(CC) $(CFLAGS) -c -o $(@:.elf=.aux.o) $(SRCDIR)/lib/auxinfo.c -DPROGNAME=\"$@\" + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/flat.lds $(filter %.o, $^) $(FLATLIBS) $(snippets) $(@:.elf=.aux.o)
Wow, weird Makefile magic, never seend that before ... but it sounds reasonable, so:
PATCH 3/3 s390x: mvpg: Add SIE mvpg test
In reply to Janosch FrankLet's also check the PEI values to make sure our VSIE implementation is correct.
--- s390x/Makefile | 2 + s390x/mvpg-sie.c | 150 ++++++++++++++++++++++++++++++++ s390x/snippets/c/mvpg-snippet.c | 33 +++++++ s390x/unittests.cfg | 3 + 4 files changed, 188 insertions(+) create mode 100644 s390x/mvpg-sie.c create mode 100644 s390x/snippets/c/mvpg-snippet.c
s390x/Makefile +2 -0
@@ -23,6 +23,7 @@ tests += $(TEST_DIR)/sie.elf tests += $(TEST_DIR)/mvpg.elf tests += $(TEST_DIR)/uv-host.elf tests += $(TEST_DIR)/edat.elf +tests += $(TEST_DIR)/mvpg-sie.elf tests_binary = $(patsubst %.elf,%.bin,$(tests)) ifneq ($(HOST_KEY_DOCUMENT),)
@@ -82,6 +83,7 @@ snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o # perquisites (=guests) for the snippet hosts. # $(TEST_DIR)/<snippet-host>.elf: snippets = $(SNIPPET_DIR)/<c/asm>/<snippet>.gbin +$(TEST_DIR)/mvpg-sie.elf: snippets = $(SNIPPET_DIR)/c/mvpg-snippet.gbin $(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o $(FLATLIBS) $(OBJCOPY) -O binary $(patsubst %.gbin,%.o,$@) $@
s390x/mvpg-sie.c new file +150 -0
150 lines not on this page · raw message
s390x/snippets/c/mvpg-snippet.c new file +33 -0
@@ -0,0 +1,33 @@ +#include <libcflat.h> + +static inline void force_exit (void ) +{ + asmvolatile (" diag 0,0,0x44\n" ); +} + +static inline int mvpg (unsigned long r0,void *dest,void *src) +{ + register unsigned long reg0 asm ("0") = r0; +int cc; + + asmvolatile (" mvpg %1,%2\n" +" ipm %0\n" +" srl %0,28" + :"=&d" (cc) :"a" (dest),"a" (src),"d" (reg0) + :"memory" ,"cc" ); +return cc; +} + +static void test_mvpg_real (void ) +{ +mvpg (0 , (void *)0x5000 , (void *)0x6000 ); +force_exit (); +} + +__attribute__((section (".text" )))int main (void ) +{ +test_mvpg_real (); +test_mvpg_real (); +test_mvpg_real (); +return 0 ; +}
s390x/unittests.cfg +3 -0
@@ -106,3 +106,6 @@ timeout = 10 [edat] file = edat.elf + +[mvpg-sie] +file = mvpg-sie.elf
Signature · 1 line
2.27.0
Re: [PATCH 3/3] s390x: mvpg: Add SIE mvpg test
Quoted 45 lines · Janosch Frank
Let's also check the PEI values to make sure our VSIE implementation is correct.
Signed-off-by Janosch Frank <[email protected]>--- s390x/Makefile | 2 + s390x/mvpg-sie.c | 150 ++++++++++++++++++++++++++++++++ s390x/snippets/c/mvpg-snippet.c | 33 +++++++ s390x/unittests.cfg | 3 + 4 files changed, 188 insertions(+)create mode 100644
s390x/mvpg-sie.c
create mode 100644s390x/snippets/c/mvpg-snippet.cs390x/Makefile +2
-0@@ -23,6 +23,7 @@ tests += $(TEST_DIR)/sie.elf tests += $(TEST_DIR)/mvpg.elf tests += $(TEST_DIR)/uv-host.elf tests += $(TEST_DIR)/edat.elf +tests += $(TEST_DIR)/mvpg-sie.elf tests_binary = $(patsubst %.elf,%.bin,$(tests)) ifneq ($(HOST_KEY_DOCUMENT),)@@ -82,6 +83,7 @@ snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o # perquisites (=guests) for the snippet hosts. # $(TEST_DIR)/<snippet-host>.elf: snippets = $(SNIPPET_DIR)/<c/asm>/<snippet>.gbin +$(TEST_DIR)/mvpg-sie.elf: snippets = $(SNIPPET_DIR)/c/mvpg-snippet.gbin $(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o $(FLATLIBS) $(OBJCOPY) -O binary $(patsubst %.gbin,%.o,$@) $@s390x/mvpg-sie.c new file +6
-0@@ -0,0 +1,150 @@ +#include <libcflat.h> +#include <asm/asm-offsets.h> +#include <asm-generic/barrier.h> +#include <asm/interrupt.h> +#include <asm/pgtable.h> +#include <mmu.h>
s390x/mvpg-sie.c line 16
Quoted 10 lines
s390x/mvpg-sie.c new file +10
-0+#include <asm/page.h> +#include <asm/facility.h> +#include <asm/mem.h> +#include <asm/sigp.h> +#include <smp.h> +#include <alloc_page.h> +#include <bitops.h> +#include <vm.h> +#include <sclp.h> +#include <sie.h>
The list of headers that get included here is rather long for this file that is rather short ... e.g. do we really need such headers like sigp.h and sclp.h here?
Quoted 11 lines
s390x/mvpg-sie.c new file +11
-0+static u8 *guest; +static u8 *guest_instr; +static struct vm vm; + +static uint8_t *src; +static uint8_t *dst; +static uint8_t *cmp; + +extern const char _binary_s390x_snippets_c_mvpg_snippet_gbin_start[]; +extern const char _binary_s390x_snippets_c_mvpg_snippet_gbin_end[]; +int binary_size;
s390x/mvpg-sie.c line 38
Quoted 10 lines
s390x/mvpg-sie.c new file +10
-0+ +static void sie (struct vm *vm) +{ +/* Reset icptcode so we don't trip below */ + vm->sblk->icptcode =0 ; + +while (vm->sblk->icptcode ==0 ) { +sie64a (vm->sblk, &vm->save_area); +if (vm->sblk->icptcode ==ICPT_VALIDITY ) +assert (0 );
assert(vm->sblk->icptcode != ICPT_VALIDITY)
?
s390x/mvpg-sie.c line 47
Quoted 9 lines
s390x/mvpg-sie.c new file +9
-0+ } + vm->save_area.guest.grs[14 ] = vm->sblk->gg14; + vm->save_area.guest.grs[15 ] = vm->sblk->gg15; +} + +static void test_mvpg_pei (void ) +{ +uint64_t **pei_dst = (uint64_t **)((uintptr_t ) vm.sblk +0xc0 ); +uint64_t **pei_src = (uint64_t **)((uintptr_t ) vm.sblk +0xc8 );
Hmm, magic values ... according to the public SA22-7095-0 from 1984, the array at offset 0xc0 is called "Interruption parameters" ... could we maybe at least use a similar name in our kvm_s390_sie_block?
Quoted 5 lines
s390x/mvpg-sie.c new file +5
-0+report_prefix_push ("pei" ); + +report_prefix_push ("src" ); +memset (dst,0 ,PAGE_SIZE ); +protect_page (src,PAGE_ENTRY_I );
s390x/mvpg-sie.c line 63
Quoted 10 lines
s390x/mvpg-sie.c new file +10
-0+sie (&vm); +report (vm.sblk->icptcode ==ICPT_PARTEXEC ,"Partial execution" ); +report ((uintptr_t )**pei_src == ((uintptr_t )vm.sblk->mso) +0x6000 +PAGE_ENTRY_I ,"PEI_SRC correct" ); +report ((uintptr_t )**pei_dst == vm.sblk->mso +0x5000 ,"PEI_DST correct" ); +unprotect_page (src,PAGE_ENTRY_I ); +report (!memcmp (cmp, dst,PAGE_SIZE ),"Destination intact" ); +/* Jump over the diag44 */ +sie (&vm); +assert (vm.sblk->icptcode ==ICPT_INST && + vm.sblk->ipa ==0x8300 && vm.sblk->ipb ==0x440000 );
Maybe add a comment about the magic values?
Quoted 99 lines
s390x/mvpg-sie.c new file +87
-0+report_prefix_pop (); + +/* Clear PEI data for next check */ +report_prefix_push ("dst" ); +memset ((uint64_t *)((uintptr_t ) vm.sblk +0xc0 ),0 ,16 ); +memset (dst,0 ,PAGE_SIZE ); +protect_page (dst,PAGE_ENTRY_I ); +sie (&vm); +report (vm.sblk->icptcode ==ICPT_PARTEXEC ,"Partial execution" ); +report ((uintptr_t )**pei_src == vm.sblk->mso +0x6000 ,"PEI_SRC correct" ); +report ((uintptr_t )**pei_dst == vm.sblk->mso +0x5000 +PAGE_ENTRY_I ,"PEI_DST correct" ); +/* Needed for the memcmp and general cleanup */ +unprotect_page (dst,PAGE_ENTRY_I ); +report (!memcmp (cmp, dst,PAGE_SIZE ),"Destination intact" ); +report_prefix_pop (); + +report_prefix_pop (); +} + +static void test_mvpg (void ) +{ +int binary_size = ((uintptr_t )_binary_s390x_snippets_c_mvpg_snippet_gbin_end - + (uintptr_t )_binary_s390x_snippets_c_mvpg_snippet_gbin_start); + +memcpy (guest, _binary_s390x_snippets_c_mvpg_snippet_gbin_start, binary_size); +memset (src,0x42 ,PAGE_SIZE ); +memset (dst,0x43 ,PAGE_SIZE ); +sie (&vm); +mb (); +report (!memcmp (src, dst,PAGE_SIZE ) && *dst ==0x42 ,"Page moved" ); +} + +static void setup_guest (void ) +{ +setup_vm (); + +/* Allocate 1MB as guest memory */ + guest =alloc_pages (8 ); +/* The first two pages are the lowcore */ + guest_instr = guest +PAGE_SIZE *2 ; + + vm.sblk =alloc_page (); + + vm.sblk->cpuflags =CPUSTAT_ZARCH |CPUSTAT_RUNNING ; + vm.sblk->prefix =0 ; +/* +* Pageable guest with the same ASCE as the test programm, but +* the guest memory 0x0 is offset to start at the allocated +* guest pages and end after 1MB. +* +* It's not pretty but faster and easier than managing guest ASCEs. +*/ + vm.sblk->mso = (u64 )guest; + vm.sblk->msl = (u64 )guest; + vm.sblk->ihcpu =0xffff ; + + vm.sblk->crycbd = (uint64_t )alloc_page (); + + vm.sblk->gpsw.addr =PAGE_SIZE *4 ; + vm.sblk->gpsw.mask =0x0000000180000000ULL ; + vm.sblk->ictl =ICTL_OPEREXC |ICTL_PINT ; +/* Enable MVPG interpretation as we want to test KVM and not ourselves */ + vm.sblk->eca =ECA_MVPGI ; + + src = guest +PAGE_SIZE *6 ; + dst = guest +PAGE_SIZE *5 ; + cmp =alloc_page (); +memset (cmp,0 ,PAGE_SIZE ); +} + +int main (void ) +{ +report_prefix_push ("mvpg-sie" ); +if (!sclp_facilities.has_sief2) { +report_skip ("SIEF2 facility unavailable" ); + goto done; + } + +setup_guest (); +test_mvpg (); +test_mvpg_pei (); + +done: +report_prefix_pop (); +return report_summary (); + +}s390x/snippets/c/mvpg-snippet.c new file +6
-0@@ -0,0 +1,33 @@ +#include <libcflat.h> + +static inline void force_exit (void ) +{ + asmvolatile (" diag 0,0,0x44\n" ); +}
s390x/snippets/c/mvpg-snippet.c line 16
Quoted 10 lines
s390x/snippets/c/mvpg-snippet.c new file +10
-0+ +static inline int mvpg (unsigned long r0,void *dest,void *src) +{ + register unsigned long reg0 asm ("0") = r0; +int cc; + + asmvolatile (" mvpg %1,%2\n" +" ipm %0\n" +" srl %0,28" + : "=&d" (cc) : "a" (dest), "a" (src), "d" (reg0)
Why is cc marked with "&" here?
Thomas
Re: [PATCH 3/3] s390x: mvpg: Add SIE mvpg test
Quoted 52 lines · Thomas Huth
Quoted 51 lines · Janosch Frank
Let's also check the PEI values to make sure our VSIE implementation is correct.
Signed-off-by Janosch Frank <[email protected]>--- s390x/Makefile | 2 + s390x/mvpg-sie.c | 150 ++++++++++++++++++++++++++++++++ s390x/snippets/c/mvpg-snippet.c | 33 +++++++ s390x/unittests.cfg | 3 + 4 files changed, 188 insertions(+)create mode 100644
s390x/mvpg-sie.c
create mode 100644s390x/snippets/c/mvpg-snippet.cs390x/Makefile +2
-0@@ -23,6 +23,7 @@ tests += $(TEST_DIR)/sie.elf tests += $(TEST_DIR)/mvpg.elf tests += $(TEST_DIR)/uv-host.elf tests += $(TEST_DIR)/edat.elf +tests += $(TEST_DIR)/mvpg-sie.elf tests_binary = $(patsubst %.elf,%.bin,$(tests)) ifneq ($(HOST_KEY_DOCUMENT),)@@ -82,6 +83,7 @@ snippet_asmlib = $(SNIPPET_DIR)/c/cstart.o # perquisites (=guests) for the snippet hosts. # $(TEST_DIR)/<snippet-host>.elf: snippets = $(SNIPPET_DIR)/<c/asm>/<snippet>.gbin +$(TEST_DIR)/mvpg-sie.elf: snippets = $(SNIPPET_DIR)/c/mvpg-snippet.gbin $(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o $(FLATLIBS) $(OBJCOPY) -O binary $(patsubst %.gbin,%.o,$@) $@s390x/mvpg-sie.c new file +12
-0@@ -0,0 +1,150 @@ +#include <libcflat.h> +#include <asm/asm-offsets.h> +#include <asm-generic/barrier.h> +#include <asm/interrupt.h> +#include <asm/pgtable.h> +#include <mmu.h> +#include <asm/page.h> +#include <asm/facility.h> +#include <asm/mem.h> +#include <asm/sigp.h> +#include <smp.h> +#include <alloc_page.h>
Quoted 8 lines
s390x/mvpg-sie.c line 16
Quoted 4 lines
s390x/mvpg-sie.c new file +4
-0+#include <bitops.h> +#include <vm.h> +#include <sclp.h> +#include <sie.h> The list of headers that get included here is rather long for this file that is rather short ... e.g. do we really need such headers like
sigp.handsclp.hhere?
sclp.h is used for the sie facility check.
I've removed asm/sigp.h, smp.h and bitops.h
Quoted 28 lines
Quoted 12 lines
s390x/mvpg-sie.c new file +11
-0+static u8 *guest; +static u8 *guest_instr; +static struct vm vm; + +static uint8_t *src; +static uint8_t *dst; +static uint8_t *cmp; + +extern const char _binary_s390x_snippets_c_mvpg_snippet_gbin_start[]; +extern const char _binary_s390x_snippets_c_mvpg_snippet_gbin_end[]; +int binary_size;s390x/mvpg-sie.c line 38
Quoted 10 lines
s390x/mvpg-sie.c new file +10
-0+ +static void sie (struct vm *vm) +{ +/* Reset icptcode so we don't trip below */ + vm->sblk->icptcode =0 ; + +while (vm->sblk->icptcode ==0 ) { +sie64a (vm->sblk, &vm->save_area); +if (vm->sblk->icptcode ==ICPT_VALIDITY ) +assert (0 );assert(vm->sblk->icptcode != ICPT_VALIDITY)?
Quoted 14 lines
s390x/mvpg-sie.c line 47
Quoted 9 lines
s390x/mvpg-sie.c new file +9
-0+ } + vm->save_area.guest.grs[14 ] = vm->sblk->gg14; + vm->save_area.guest.grs[15 ] = vm->sblk->gg15; +} + +static void test_mvpg_pei (void ) +{ +uint64_t **pei_dst = (uint64_t **)((uintptr_t ) vm.sblk +0xc0 ); +uint64_t **pei_src = (uint64_t **)((uintptr_t ) vm.sblk +0xc8 );Hmm, magic values ... according to the public SA22-7095-0 from 1984, the array at offset
0xc0is called "Interruption parameters" ... could we maybe at least use a similar name in ourkvm_s390_sie_block?
That's very much a thing that moved over from KVM and the reason for that is that the interruption parameters contain a lot of things especially when PV is added into the mix.
We can now either choose to make the interruption parameters an interesing looking union or we define the offsets properly. I'm also not too happy about the VSIE PEI code using "u64 *pei_block = &vsie_page->scb_o->mcic;" for that.
#define SBLK_OFF_PEI_DST 0xc0
#define SBLK_OFF_PEI_SRC 0xc8
Quoted 13 lines
s390x/mvpg-sie.c line 60
Quoted 13 lines
s390x/mvpg-sie.c new file +12
-0+report_prefix_push ("pei" ); + +report_prefix_push ("src" ); +memset (dst,0 ,PAGE_SIZE ); +protect_page (src,PAGE_ENTRY_I ); +sie (&vm); +report (vm.sblk->icptcode ==ICPT_PARTEXEC ,"Partial execution" ); +report ((uintptr_t )**pei_src == ((uintptr_t )vm.sblk->mso) +0x6000 +PAGE_ENTRY_I ,"PEI_SRC correct" ); +report ((uintptr_t )**pei_dst == vm.sblk->mso +0x5000 ,"PEI_DST correct" ); +unprotect_page (src,PAGE_ENTRY_I ); +report (!memcmp (cmp, dst,PAGE_SIZE ),"Destination intact" ); +/* Jump over the diag44 */
How about:
/*
* We need to execute the diag44 which is used as a blocker behind the
* mvpg. It makes sure we fail the tests above if the mvpg wouldn't
* have intercepted.
*/
Quoted 1 line
s390x/mvpg-sie.c line 59
Quoted 1 line
s390x/mvpg-sie.c new file +1
-0+sie (&vm);
And:
/* Make sure we intercepted for the diag44 and noting else */
Quoted 4 lines
s390x/mvpg-sie.c line 63
Quoted 2 lines
s390x/mvpg-sie.c new file +2
-0+assert (vm.sblk->icptcode ==ICPT_INST && + vm.sblk->ipa ==0x8300 && vm.sblk->ipb ==0x440000 );Maybe add a comment about the magic values?
Since these will be used extensively in the future I think it might make sense to introduce constants.
#define SIE_ICPT_IPA_DIAG 0x8300
Quoted 104 lines
Quoted 104 lines
s390x/mvpg-sie.c new file +87
-0+report_prefix_pop (); + +/* Clear PEI data for next check */ +report_prefix_push ("dst" ); +memset ((uint64_t *)((uintptr_t ) vm.sblk +0xc0 ),0 ,16 ); +memset (dst,0 ,PAGE_SIZE ); +protect_page (dst,PAGE_ENTRY_I ); +sie (&vm); +report (vm.sblk->icptcode ==ICPT_PARTEXEC ,"Partial execution" ); +report ((uintptr_t )**pei_src == vm.sblk->mso +0x6000 ,"PEI_SRC correct" ); +report ((uintptr_t )**pei_dst == vm.sblk->mso +0x5000 +PAGE_ENTRY_I ,"PEI_DST correct" ); +/* Needed for the memcmp and general cleanup */ +unprotect_page (dst,PAGE_ENTRY_I ); +report (!memcmp (cmp, dst,PAGE_SIZE ),"Destination intact" ); +report_prefix_pop (); + +report_prefix_pop (); +} + +static void test_mvpg (void ) +{ +int binary_size = ((uintptr_t )_binary_s390x_snippets_c_mvpg_snippet_gbin_end - + (uintptr_t )_binary_s390x_snippets_c_mvpg_snippet_gbin_start); + +memcpy (guest, _binary_s390x_snippets_c_mvpg_snippet_gbin_start, binary_size); +memset (src,0x42 ,PAGE_SIZE ); +memset (dst,0x43 ,PAGE_SIZE ); +sie (&vm); +mb (); +report (!memcmp (src, dst,PAGE_SIZE ) && *dst ==0x42 ,"Page moved" ); +} + +static void setup_guest (void ) +{ +setup_vm (); + +/* Allocate 1MB as guest memory */ + guest =alloc_pages (8 ); +/* The first two pages are the lowcore */ + guest_instr = guest +PAGE_SIZE *2 ; + + vm.sblk =alloc_page (); + + vm.sblk->cpuflags =CPUSTAT_ZARCH |CPUSTAT_RUNNING ; + vm.sblk->prefix =0 ; +/* +* Pageable guest with the same ASCE as the test programm, but +* the guest memory 0x0 is offset to start at the allocated +* guest pages and end after 1MB. +* +* It's not pretty but faster and easier than managing guest ASCEs. +*/ + vm.sblk->mso = (u64 )guest; + vm.sblk->msl = (u64 )guest; + vm.sblk->ihcpu =0xffff ; + + vm.sblk->crycbd = (uint64_t )alloc_page (); + + vm.sblk->gpsw.addr =PAGE_SIZE *4 ; + vm.sblk->gpsw.mask =0x0000000180000000ULL ; + vm.sblk->ictl =ICTL_OPEREXC |ICTL_PINT ; +/* Enable MVPG interpretation as we want to test KVM and not ourselves */ + vm.sblk->eca =ECA_MVPGI ; + + src = guest +PAGE_SIZE *6 ; + dst = guest +PAGE_SIZE *5 ; + cmp =alloc_page (); +memset (cmp,0 ,PAGE_SIZE ); +} + +int main (void ) +{ +report_prefix_push ("mvpg-sie" ); +if (!sclp_facilities.has_sief2) { +report_skip ("SIEF2 facility unavailable" ); + goto done; + } + +setup_guest (); +test_mvpg (); +test_mvpg_pei (); + +done: +report_prefix_pop (); +return report_summary (); + +}s390x/snippets/c/mvpg-snippet.c new file +10
-0@@ -0,0 +1,33 @@ +#include <libcflat.h> + +static inline void force_exit (void ) +{ + asmvolatile (" diag 0,0,0x44\n" ); +} + +static inline int mvpg (unsigned long r0,void *dest,void *src) +{ + register unsigned long reg0 asm ("0") = r0;
Quoted 8 lines
s390x/snippets/c/mvpg-snippet.c line 16
Quoted 6 lines
s390x/snippets/c/mvpg-snippet.c new file +6
-0+int cc; + + asmvolatile (" mvpg %1,%2\n" +" ipm %0\n" +" srl %0,28" + : "=&d" (cc) : "a" (dest), "a" (src), "d" (reg0)Why is cc marked with "&" here?
I copied that from s390x/mvpg.c
Quoted 3 lines
Thomas