s390x: PV fixups
cover s390x: PV fixups
A small set of patches that clean up the PV snippet handling.
Janosch Frank (6):
s390x: snippets: asm: Add a macro to write an exception PSW
s390x: MAKEFILE: Use $< instead of pathsubst
s390x: Add a linker script to assembly snippets
lib: s390x: sie: Improve validity handling and make it vm specific
lib: s390x: Use a new asce for each PV guest
lib: s390x: sie: Properly populate SCA
lib/s390x/asm-offsets.c | 2 ++ lib/s390x/sie.c | 36 +++++++++++++------- lib/s390x/sie.h | 43 ++++++++++++++++++++++-- lib/s390x/snippet.h | 3 +- lib/s390x/uv.c | 35 +++++++++++++++++-- lib/s390x/uv.h | 5 ++- s390x/Makefile | 18 +++++++--- s390x/cpu.S | 6 ++++ s390x/mvpg-sie.c | 2 +- s390x/pv-diags.c | 6 ++-- s390x/snippets/asm/macros.S | 28 +++++++++++++++ s390x/snippets/asm/snippet-pv-diag-288.S | 4 +-- s390x/snippets/asm/snippet-pv-diag-500.S | 6 ++-- 13 files changed, 157 insertions(+), 37 deletions(-) create mode 100644 s390x/snippets/asm/macros.S
Signature · 1 line
2.34.1
PATCH 1/6 s390x: snippets: asm: Add a macro to write an exception PSW
Setting exception new PSWs is commonly needed so let's add a macro for that.
--- s390x/snippets/asm/macros.S | 28 ++++++++++++++++++++++++ s390x/snippets/asm/snippet-pv-diag-288.S | 4 ++-- s390x/snippets/asm/snippet-pv-diag-500.S | 6 ++--- 3 files changed, 32 insertions(+), 6 deletions(-) create mode 100644 s390x/snippets/asm/macros.S
s390x/snippets/asm/macros.S new file +28 -0
@@ -0,0 +1,28 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Commonly used assembly macros + * + * Copyright (c) 2022 IBM Corp + * + * Authors: + * Janosch Frank <[email protected]> + */ +#include <asm/asm-offsets.h> + +/* + * Writes a PSW to addr_psw, useful for exception PSWs in lowcore + * + * reg is the scratch register used for temporary storage, it's NOT restored + * The psw address part is defined via psw_new_addr + * The psw mask part is always 64 bit + */ +.macro SET_PSW_NEW_ADDR reg, psw_new_addr, addr_psw +larl \reg, psw_mask_64 +stg \reg, \addr_psw +larl \reg, \psw_new_addr +stg \reg, \addr_psw + 8 +.endm + +.section .rodata +psw_mask_64: + .quad 0x0000000180000000
s390x/snippets/asm/snippet-pv-diag-288.S +2 -2
@@ -8,6 +8,7 @@ * Janosch Frank <[email protected]> */ #include <asm/asm-offsets.h> +#include "macros.S" .section .text /* Clean and pre-load registers that are used for diag 288 */
@@ -19,8 +20,7 @@ lghi %r1, 2 lghi %r2, 3 /* Let's jump to the pgm exit label on a PGM */ -larl %r4, exit_pgm -stg %r4, GEN_LC_PGM_NEW_PSW + 8 +SET_PSW_NEW_ADDR 4, exit_pgm, GEN_LC_PGM_NEW_PSW /* Execute the diag288 */ diag %r0, %r2, 0x288
s390x/snippets/asm/snippet-pv-diag-500.S +2 -4
@@ -8,6 +8,7 @@ * Janosch Frank <[email protected]> */ #include <asm/asm-offsets.h> +#include "macros.S" .section .text /* Clean and pre-load registers that are used for diag 500 */
@@ -21,10 +22,7 @@ lghi %r3, 3 lghi %r4, 4 /* Let's jump to the next label on a PGM */ -xgr %r5, %r5 -stg %r5, GEN_LC_PGM_NEW_PSW -larl %r5, next -stg %r5, GEN_LC_PGM_NEW_PSW + 8 +SET_PSW_NEW_ADDR 5, next, GEN_LC_PGM_NEW_PSW /* Execute the diag500 */ diag 0, 0, 0x500
Signature · 1 line
2.34.1
Re: [kvm-unit-tests PATCH 1/6] s390x: snippets: asm: Add a macro to write an exception PSW
Quoting Janosch Frank (2022-07-29 10:26:28)
[...]
Quoted 13 lines
s390x/snippets/asm/snippet-pv-diag-500.S +1
-0@@ -8,6 +8,7 @@ * Janosch Frank <[email protected]> */ #include <asm/asm-offsets.h> +#include "macros.S" .section .text /* Clean and pre-load registers that are used for diag 500 */
s390x/snippets/asm/snippet-pv-diag-500.S lghi %r3, 3 · line 25
Quoted 5 lines
s390x/snippets/asm/snippet-pv-diag-500.S +0
-2@@ -21,10 +22,7 @@ lghi %r3, 3 lghi %r4, 4 /* Let's jump to the next label on a PGM */ -xgr %r5, %r5 -stg %r5, GEN_LC_PGM_NEW_PSW
So previously the PSW mask was zero and hence we had 24-bit addressing, no? Now, we have bits 31 and 32 one and hence 64 bit addressing.
I guess 24-bit addressing is not appropriate here (or at least doesn't matter too much), so I guess this is intended, isn't it?
Re: [kvm-unit-tests PATCH 1/6] s390x: snippets: asm: Add a macro to write an exception PSW
Quoted 17 lines · Nico Boehr
Quoting Janosch Frank (2022-07-29 10:26:28)
[...]Quoted 15 lines
s390x/snippets/asm/snippet-pv-diag-500.S +1
-0@@ -8,6 +8,7 @@ * Janosch Frank <[email protected]> */ #include <asm/asm-offsets.h> +#include "macros.S" .section .text /* Clean and pre-load registers that are used for diag 500 */@@ -21,10 +22,7 @@ lghi %r3, 3 lghi %r4, 4
Quoted 5 lines
s390x/snippets/asm/snippet-pv-diag-500.S lghi %r3, 3 · line 25
Quoted 3 lines
s390x/snippets/asm/snippet-pv-diag-500.S +0
-2/* Let's jump to the next label on a PGM */ -xgr %r5, %r5 -stg %r5, GEN_LC_PGM_NEW_PSWSo previously the PSW mask was zero and hence we had 24-bit addressing, no? Now, we have bits 31 and 32 one and hence 64 bit addressing.
Yes
Also the linker script patch will exchange the mask for an invalid one so we need to replace both the mask and the address.
Quoted 2 lines
I guess 24-bit addressing is not appropriate here (or at least doesn't matter too much), so I guess this is intended, isn't it?
Claudio complained about the addressing change so I moved it over to full 64 bit.
Nico Boehr
Reviewed-byRe: [kvm-unit-tests PATCH 1/6] s390x: snippets: asm: Add a macro to write an exception PSW
Quoting Janosch Frank (2022-08-02 09:56:12)
Quoted 24 lines · Nico Boehr
Quoted 17 lines
Quoting Janosch Frank (2022-07-29 10:26:28)
[...]Quoted 15 lines
s390x/snippets/asm/snippet-pv-diag-500.S +1
-0@@ -8,6 +8,7 @@ * Janosch Frank <[email protected]> */ #include <asm/asm-offsets.h> +#include "macros.S" .section .text /* Clean and pre-load registers that are used for diag 500 */@@ -21,10 +22,7 @@ lghi %r3, 3 lghi %r4, 4Quoted 5 lines
s390x/snippets/asm/snippet-pv-diag-500.S lghi %r3, 3 · line 25
Quoted 3 lines
s390x/snippets/asm/snippet-pv-diag-500.S +0
-2/* Let's jump to the next label on a PGM */ -xgr %r5, %r5 -stg %r5, GEN_LC_PGM_NEW_PSWSo previously the PSW mask was zero and hence we had 24-bit addressing, no? Now, we have bits 31 and 32 one and hence 64 bit addressing.
Quoted 6 lines
Yes
Also the linker script patch will exchange the mask for an invalid one so we need to replace both the mask and the address.Quoted 2 lines
I guess 24-bit addressing is not appropriate here (or at least doesn't matter too much), so I guess this is intended, isn't it?
Allrighty, then:
PATCH 2/6 s390x: MAKEFILE: Use $< instead of pathsubst
In reply to Janosch FrankNo need to mangle strings if we already have the value at hand.
--- s390x/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
s390x/Makefile +1 -1
@@ -134,7 +134,7 @@ $(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o truncate -s '%4096' $@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS) - $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/c/flat.lds $(patsubst %.gbin,%.o,$@) $(snippet_lib) $(FLATLIBS) + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/c/flat.lds $< $(snippet_lib) $(FLATLIBS) $(OBJCOPY) -O binary -j ".rodata" -j ".lowcore" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $@ $@ truncate -s '%4096' $@
Signature · 1 line
2.34.1
Re: [kvm-unit-tests PATCH 2/6] s390x: MAKEFILE: Use $< instead of pathsubst
Quoted 3 lines · Janosch Frank
No need to mangle strings if we already have the value at hand.
Signed-off-by Janosch Frank <[email protected]>
Quoted 17 lines
--- s390x/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)s390x/Makefile +1
-1@@ -134,7 +134,7 @@ $(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o truncate -s '%4096' $@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS) - $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/c/flat.lds $(patsubst %.gbin,%.o,$@) $(snippet_lib) $(FLATLIBS) + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/c/flat.lds $< $(snippet_lib) $(FLATLIBS) $(OBJCOPY) -O binary -j ".rodata" -j ".lowcore" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $@ $@ truncate -s '%4096' $@
Re: [kvm-unit-tests PATCH 2/6] s390x: MAKEFILE: Use $< instead of pathsubst
In reply to Janosch FrankQuoting Janosch Frank (2022-07-29 10:26:29)
Quoted 3 lines
No need to mangle strings if we already have the value at hand.
Signed-off-by Janosch Frank <[email protected]>
PATCH 3/6 s390x: Add a linker script to assembly snippets
In reply to Janosch FrankA linker script has a few benefits:
- We can easily define a lowcore and load the snippet from 0x0 instead of 0x4000 which makes asm snippets behave like c snippets
- We can easily define an invalid PGM new PSW to ensure an exit on a guest PGM
As we gain another step and file extension by linking, a comment explains which file extensions are generated and why.
--- lib/s390x/snippet.h | 3 +-- s390x/Makefile | 16 +++++++++++++--- s390x/mvpg-sie.c | 2 +- s390x/pv-diags.c | 6 +++--- 4 files changed, 18 insertions(+), 9 deletions(-)
lib/s390x/snippet.h +1 -2
@@ -32,8 +32,7 @@#define SNIPPET_PV_TWEAK0 0x42UL#define SNIPPET_PV_TWEAK1 0UL -#define SNIPPET_OFF_C 0 -#define SNIPPET_OFF_ASM 0x4000 +#define SNIPPET_UNPACK_OFF 0/*
s390x/Makefile +13 -3
@@ -122,6 +122,13 @@ else snippet-hdr-obj = endif +# Each snippet will generate the following files (in order): \ + *.o is a snippet that has been compiled \ + *.ol is a snippet that has been linked \ + *.gbin is a snippet that has been converted to binary \ + *.gobj is the final format after converting the binary into a elf file again, \ + it will be linked into the tests + # the asm/c snippets %.o have additional generated files as dependencies $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) $(CC) $(CFLAGS) -c -nostdlib -o $@ $<
@@ -129,8 +136,11 @@ $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) $(SNIPPET_DIR)/c/%.o: $(SNIPPET_DIR)/c/%.c $(asm-offsets) $(CC) $(CFLAGS) -c -nostdlib -o $@ $< -$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o - $(OBJCOPY) -O binary -j ".rodata" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $(patsubst %.gbin,%.o,$@) $@ +$(SNIPPET_DIR)/asm/%.ol: $(SNIPPET_DIR)/asm/%.o + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/asm/flat.lds $< + +$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.ol + $(OBJCOPY) -O binary -j ".rodata" -j ".lowcore" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $< $@ truncate -s '%4096' $@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS)
@@ -139,7 +149,7 @@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS) truncate -s '%4096' $@ $(SNIPPET_DIR)/asm/%.hdr: $(SNIPPET_DIR)/asm/%.gbin $(HOST_KEY_DOCUMENT) - $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x4000,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@ + $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x0,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@ $(SNIPPET_DIR)/c/%.hdr: $(SNIPPET_DIR)/c/%.gbin $(HOST_KEY_DOCUMENT) $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x0,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@
s390x/mvpg-sie.c +1 -1
@@ -87,7 +87,7 @@ static void setup_guest(void)snippet_setup_guest (&vm, false);snippet_init (&vm,SNIPPET_NAME_START (c, mvpg_snippet), -SNIPPET_LEN (c, mvpg_snippet),SNIPPET_OFF_C ); +SNIPPET_LEN (c, mvpg_snippet),SNIPPET_UNPACK_OFF );/* Enable MVPG interpretation as we want to test KVM and not ourselves */ vm.sblk->eca =ECA_MVPGI ;
s390x/pv-diags.c +3 -3
@@ -28,7 +28,7 @@ static void test_diag_500(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_500), SNIPPET_HDR_START(asm, snippet_pv_diag_500), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );sie (&vm); report(vm.sblk->icptcode == ICPT_PV_INSTR && vm.sblk->ipa == 0x8302 &&
@@ -83,7 +83,7 @@ static void test_diag_288(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_288), SNIPPET_HDR_START(asm, snippet_pv_diag_288), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );sie (&vm); report(vm.sblk->icptcode == ICPT_PV_INSTR && vm.sblk->ipa == 0x8302 &&
@@ -124,7 +124,7 @@ static void test_diag_yield(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_yield), SNIPPET_HDR_START(asm, snippet_pv_diag_yield), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );/* 0x44 */ report_prefix_push ("0x44" );
Signature · 1 line
2.34.1
Re: [kvm-unit-tests PATCH 3/6] s390x: Add a linker script to assembly snippets
Quoted 50 lines · Janosch Frank
A linker script has a few benefits:
- We can easily define a lowcore and load the snippet from0x0instead of0x4000which makes asm snippets behave like c snippets
- We can easily define an invalid PGM new PSW to ensure an exit on a guest PGMAs we gain another step and file extension by linking, a comment explains which file extensions are generated and why.
Signed-off-by Janosch Frank <[email protected]>--- lib/s390x/snippet.h | 3 +-- s390x/Makefile | 16 +++++++++++++--- s390x/mvpg-sie.c | 2 +- s390x/pv-diags.c | 6 +++--- 4 files changed, 18 insertions(+), 9 deletions(-)lib/s390x/snippet.h +1
-2@@ -32,8 +32,7 @@#define SNIPPET_PV_TWEAK0 0x42UL#define SNIPPET_PV_TWEAK1 0UL -#define SNIPPET_OFF_C 0 -#define SNIPPET_OFF_ASM 0x4000 +#define SNIPPET_UNPACK_OFF 0/* s390x/Makefile +7
-0@@ -122,6 +122,13 @@ else snippet-hdr-obj = endif +# Each snippet will generate the following files (in order): \ + *.o is a snippet that has been compiled \ + *.ol is a snippet that has been linked \ + *.gbin is a snippet that has been converted to binary \ + *.gobj is the final format after converting the binary into a elf file again, \ + it will be linked into the tests + # the asm/c snippets %.o have additional generated files as dependencies $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) $(CC) $(CFLAGS) -c -nostdlib -o $@ $<
s390x/Makefile $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) · line 140
Quoted 7 lines
s390x/Makefile +2
-2@@ -129,8 +136,11 @@ $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) $(SNIPPET_DIR)/c/%.o: $(SNIPPET_DIR)/c/%.c $(asm-offsets) $(CC) $(CFLAGS) -c -nostdlib -o $@ $< -$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o - $(OBJCOPY) -O binary -j ".rodata" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $(patsubst %.gbin,%.o,$@) $@ +$(SNIPPET_DIR)/asm/%.ol: $(SNIPPET_DIR)/asm/%.o + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/asm/flat.lds $<
I think you forgot to include that linker script in the patch. Or did I miss anything?
you can also use the $(SNIPPET_DIR) variable instead of searching for that directory yourself.
Quoted 51 lines
s390x/Makefile +4
-1+ +$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.ol + $(OBJCOPY) -O binary -j ".rodata" -j ".lowcore" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $< $@ truncate -s '%4096' $@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS)@@ -139,7 +149,7 @@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS) truncate -s '%4096' $@ $(SNIPPET_DIR)/asm/%.hdr: $(SNIPPET_DIR)/asm/%.gbin $(HOST_KEY_DOCUMENT) - $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x4000,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@ + $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x0,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@ $(SNIPPET_DIR)/c/%.hdr: $(SNIPPET_DIR)/c/%.gbin $(HOST_KEY_DOCUMENT) $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x0,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@s390x/mvpg-sie.c +1
-1@@ -87,7 +87,7 @@ static void setup_guest(void)snippet_setup_guest (&vm, false);snippet_init (&vm,SNIPPET_NAME_START (c, mvpg_snippet), -SNIPPET_LEN (c, mvpg_snippet),SNIPPET_OFF_C ); +SNIPPET_LEN (c, mvpg_snippet),SNIPPET_UNPACK_OFF );/* Enable MVPG interpretation as we want to test KVM and not ourselves */ vm.sblk->eca =ECA_MVPGI ;s390x/pv-diags.c +2
-2@@ -28,7 +28,7 @@ static void test_diag_500(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_500), SNIPPET_HDR_START(asm, snippet_pv_diag_500), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );sie (&vm); report(vm.sblk->icptcode == ICPT_PV_INSTR && vm.sblk->ipa == 0x8302 &&@@ -83,7 +83,7 @@ static void test_diag_288(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_288), SNIPPET_HDR_START(asm, snippet_pv_diag_288), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );sie (&vm); report(vm.sblk->icptcode == ICPT_PV_INSTR && vm.sblk->ipa == 0x8302 &&
s390x/pv-diags.c static void test_diag_yield(void) · line 130
Quoted 8 lines
s390x/pv-diags.c +1
-1@@ -124,7 +124,7 @@ static void test_diag_yield(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_yield), SNIPPET_HDR_START(asm, snippet_pv_diag_yield), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );/* 0x44 */ report_prefix_push ("0x44" );
Steffen Eiden
Re: [kvm-unit-tests PATCH 3/6] s390x: Add a linker script to assembly snippets
Quoted 64 lines · Steffen Eiden
Quoted 50 lines · Janosch Frank
A linker script has a few benefits:
- We can easily define a lowcore and load the snippet from0x0instead of0x4000which makes asm snippets behave like c snippets
- We can easily define an invalid PGM new PSW to ensure an exit on a guest PGMAs we gain another step and file extension by linking, a comment explains which file extensions are generated and why.
Signed-off-by Janosch Frank <[email protected]>--- lib/s390x/snippet.h | 3 +-- s390x/Makefile | 16 +++++++++++++--- s390x/mvpg-sie.c | 2 +- s390x/pv-diags.c | 6 +++--- 4 files changed, 18 insertions(+), 9 deletions(-)lib/s390x/snippet.h +1
-2@@ -32,8 +32,7 @@#define SNIPPET_PV_TWEAK0 0x42UL#define SNIPPET_PV_TWEAK1 0UL -#define SNIPPET_OFF_C 0 -#define SNIPPET_OFF_ASM 0x4000 +#define SNIPPET_UNPACK_OFF 0/* s390x/Makefile +7
-0@@ -122,6 +122,13 @@ else snippet-hdr-obj = endif +# Each snippet will generate the following files (in order): \ + *.o is a snippet that has been compiled \ + *.ol is a snippet that has been linked \ + *.gbin is a snippet that has been converted to binary \ + *.gobj is the final format after converting the binary into a elf file again, \ + it will be linked into the tests + # the asm/c snippets %.o have additional generated files as dependencies $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) $(CC) $(CFLAGS) -c -nostdlib -o $@ $<s390x/Makefile $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) · line 140
Quoted 7 lines
s390x/Makefile +2
-2@@ -129,8 +136,11 @@ $(SNIPPET_DIR)/asm/%.o: $(SNIPPET_DIR)/asm/%.S $(asm-offsets) $(SNIPPET_DIR)/c/%.o: $(SNIPPET_DIR)/c/%.c $(asm-offsets) $(CC) $(CFLAGS) -c -nostdlib -o $@ $< -$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.o - $(OBJCOPY) -O binary -j ".rodata" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $(patsubst %.gbin,%.o,$@) $@ +$(SNIPPET_DIR)/asm/%.ol: $(SNIPPET_DIR)/asm/%.o + $(CC) $(LDFLAGS) -o $@ -T $(SRCDIR)/s390x/snippets/asm/flat.lds $<I think you forgot to include that linker script in the patch. Or did I miss anything?
Quoted 2 lines
you can also use the $(SNIPPET_DIR) variable instead of searching for that directory yourself.
Right on both cases, I'll fix that...
Quoted 62 lines
Quoted 52 lines
s390x/Makefile +4
-1+ +$(SNIPPET_DIR)/asm/%.gbin: $(SNIPPET_DIR)/asm/%.ol + $(OBJCOPY) -O binary -j ".rodata" -j ".lowcore" -j ".text" -j ".data" -j ".bss" --set-section-flags .bss=alloc,load,contents $< $@ truncate -s '%4096' $@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS)@@ -139,7 +149,7 @@ $(SNIPPET_DIR)/c/%.gbin: $(SNIPPET_DIR)/c/%.o $(snippet_lib) $(FLATLIBS) truncate -s '%4096' $@ $(SNIPPET_DIR)/asm/%.hdr: $(SNIPPET_DIR)/asm/%.gbin $(HOST_KEY_DOCUMENT) - $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x4000,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@ + $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x0,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@ $(SNIPPET_DIR)/c/%.hdr: $(SNIPPET_DIR)/c/%.gbin $(HOST_KEY_DOCUMENT) $(GEN_SE_HEADER) -k $(HOST_KEY_DOCUMENT) -c $<,0x0,0x00000000000000420000000000000000 --psw-addr 0x4000 -o $@s390x/mvpg-sie.c +1
-1@@ -87,7 +87,7 @@ static void setup_guest(void)snippet_setup_guest (&vm, false);snippet_init (&vm,SNIPPET_NAME_START (c, mvpg_snippet), -SNIPPET_LEN (c, mvpg_snippet),SNIPPET_OFF_C ); +SNIPPET_LEN (c, mvpg_snippet),SNIPPET_UNPACK_OFF );/* Enable MVPG interpretation as we want to test KVM and not ourselves */ vm.sblk->eca =ECA_MVPGI ;s390x/pv-diags.c +2
-2@@ -28,7 +28,7 @@ static void test_diag_500(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_500), SNIPPET_HDR_START(asm, snippet_pv_diag_500), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );sie (&vm); report(vm.sblk->icptcode == ICPT_PV_INSTR && vm.sblk->ipa == 0x8302 &&@@ -83,7 +83,7 @@ static void test_diag_288(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_288), SNIPPET_HDR_START(asm, snippet_pv_diag_288), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );sie (&vm); report(vm.sblk->icptcode == ICPT_PV_INSTR && vm.sblk->ipa == 0x8302 &&s390x/pv-diags.c static void test_diag_yield(void) · line 130
Quoted 8 lines
s390x/pv-diags.c +1
-1@@ -124,7 +124,7 @@ static void test_diag_yield(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_yield), SNIPPET_HDR_START(asm, snippet_pv_diag_yield), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );/* 0x44 */ report_prefix_push ("0x44" );Steffen Eiden
Re: [kvm-unit-tests PATCH 3/6] s390x: Add a linker script to assembly snippets
In reply to Janosch FrankQuoting Janosch Frank (2022-07-29 10:26:30)
[...]
s390x/pv-diags.c static void test_diag_500(void) · line 31
Quoted 10 lines
s390x/pv-diags.c +1
-1@@ -28,7 +28,7 @@ static void test_diag_500(void) snippet_pv_init(&vm, SNIPPET_NAME_START(asm, snippet_pv_diag_500), SNIPPET_HDR_START(asm, snippet_pv_diag_500), - size_gbin, size_hdr,SNIPPET_OFF_ASM ); + size_gbin, size_hdr,SNIPPET_UNPACK_OFF );
Does it even make sense to pass SNIPPET_UNPACK_OFF? Since this is now the same for C and ASM we could completely get rid of this argument, couldn't we?
PATCH 4/6 lib: s390x: sie: Improve validity handling and make it vm specific
In reply to Janosch FrankThe current library doesn't support running multiple vms at once as it stores the validity once and not per vm. Let's move the validity handling into the vm and introduce a new function to retrieve the vir.
--- lib/s390x/sie.c | 26 +++++++++++++------------- lib/s390x/sie.h | 6 ++++-- 2 files changed, 17 insertions(+), 15 deletions(-)
lib/s390x/sie.c +13 -13
@@ -15,19 +15,21 @@#include <libcflat.h> #include <alloc_page.h> -static bool validity_expected; -static uint16_t vir;/* Validity interception reason */ - -void sie_expect_validity (void ) +void sie_expect_validity (struct vm *vm) { - validity_expected = true; - vir =0 ; + vm->validity_expected = true; } -void sie_check_validity (uint16_t vir_exp) +uint16_t sie_get_validity (struct vm *vm) { +return vm->sblk->ipb >>16 ; +} + +void sie_check_validity (struct vm *vm,uint16_t vir_exp) +{ +uint16_t vir =sie_get_validity (vm); +report (vir_exp == vir,"VALIDITY: %x" , vir); - vir =0 ; }void sie_handle_validity (struct vm *vm)
@@ -35,11 +37,9 @@ void sie_handle_validity(struct vm *vm)if (vm->sblk->icptcode !=ICPT_VALIDITY )return ; - vir = vm->sblk->ipb >>16 ; - -if (!validity_expected) -report_abort ("VALIDITY: %x" , vir); - validity_expected = false; +if (!vm->validity_expected) +report_abort ("VALIDITY: %x" ,sie_get_validity (vm)); + vm->validity_expected = false; }void sie (struct vm *vm)
lib/s390x/sie.h +4 -2
@@ -233,14 +233,16 @@ struct vm {struct vm_uv uv;/* PV UV information */ /* Ptr to first guest page */ uint8_t *guest_mem; +bool validity_expected; };extern void sie_entry (void );extern void sie_exit (void );extern void sie64a (struct kvm_s390_sie_block *sblk,struct vm_save_area *save_area);void sie (struct vm *vm); -void sie_expect_validity (void ); -void sie_check_validity (uint16_t vir_exp); +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);void sie_guest_sca_create (struct vm *vm);void sie_guest_create (struct vm *vm,uint64_t guest_mem,uint64_t guest_mem_len);
Signature · 1 line
2.34.1
Re: [kvm-unit-tests PATCH 4/6] lib: s390x: sie: Improve validity handling and make it vm specific
Quoted 5 lines · Janosch Frank
The current library doesn't support running multiple vms at once as it stores the validity once and not per vm. Let's move the validity handling into the vm and introduce a new function to retrieve the vir.
Signed-off-by Janosch Frank <[email protected]>
Re: [kvm-unit-tests PATCH 4/6] lib: s390x: sie: Improve validity handling and make it vm specific
In reply to Janosch FrankQuoting Janosch Frank (2022-07-29 10:26:31)
Quoted 5 lines
The current library doesn't support running multiple vms at once as it stores the validity once and not per vm. Let's move the validity handling into the vm and introduce a new function to retrieve the vir.
Signed-off-by Janosch Frank <[email protected]>
One tiny suggestion below.
[...]
Quoted 14 lines
lib/s390x/sie.c +1
-4@@ -15,19 +15,21 @@#include <libcflat.h> #include <alloc_page.h> -static bool validity_expected; -static uint16_t vir;/* Validity interception reason */ - -void sie_expect_validity (void ) +void sie_expect_validity (struct vm *vm) {
lib/s390x/sie.c line 26
Quoted 10 lines
lib/s390x/sie.c +4
-3- validity_expected = true; - vir =0 ; + vm->validity_expected = true; } -void sie_check_validity (uint16_t vir_exp) +uint16_t sie_get_validity (struct vm *vm) { +return vm->sblk->ipb >>16 ; +}
Since one should only call this function when we had a validity, maybe add:
assert (vm->sblk->icptcode == ICPT_VALIDITY);
PATCH 5/6 lib: s390x: Use a new asce for each PV guest
In reply to Janosch FrankEvery PV guest needs its own ASCE so let's copy the topmost table designated by CR1 to create a new ASCE for the PV guest. Before and after SIE we now need to switch ASCEs to and from the PV guest / test ASCE. The SIE assembly function does that automatically.
--- lib/s390x/asm-offsets.c | 2 ++ lib/s390x/sie.c | 2 ++ lib/s390x/sie.h | 2 ++ lib/s390x/uv.c | 35 +++++++++++++++++++++++++++++++++-- lib/s390x/uv.h | 5 ++--- s390x/cpu.S | 6 ++++++ 6 files changed, 47 insertions(+), 5 deletions(-)
lib/s390x/asm-offsets.c +2 -0
@@ -75,9 +75,11 @@ int main(void)OFFSET (SIE_SAVEAREA_HOST_GRS , vm_save_area, host.grs[0 ]);OFFSET (SIE_SAVEAREA_HOST_FPRS , vm_save_area, host.fprs[0 ]);OFFSET (SIE_SAVEAREA_HOST_FPC , vm_save_area, host.fpc); +OFFSET (SIE_SAVEAREA_HOST_ASCE , vm_save_area, host.asce);OFFSET (SIE_SAVEAREA_GUEST_GRS , vm_save_area, guest.grs[0 ]);OFFSET (SIE_SAVEAREA_GUEST_FPRS , vm_save_area, guest.fprs[0 ]);OFFSET (SIE_SAVEAREA_GUEST_FPC , vm_save_area, guest.fpc); +OFFSET (SIE_SAVEAREA_GUEST_ASCE , vm_save_area, guest.asce);OFFSET (STACK_FRAME_INT_BACKCHAIN , stack_frame_int, back_chain);OFFSET (STACK_FRAME_INT_FPC , stack_frame_int, fpc);OFFSET (STACK_FRAME_INT_FPRS , stack_frame_int, fprs);
lib/s390x/sie.c +2 -0
@@ -84,6 +84,8 @@ void sie_guest_create(struct vm *vm, uint64_t guest_mem, uint64_t guest_mem_len)/* Guest memory chunks are always 1MB */ assert (!(guest_mem_len & ~HPAGE_MASK )); +/* For non-PV guests we re-use the host's ASCE for ease of use */ + vm->save_area.guest.asce =stctg (1 );/* Currently MSO/MSL is the easiest option */ vm->sblk->mso = (uint64_t )guest_mem; vm->sblk->msl = (uint64_t )guest_mem + ((guest_mem_len -1 ) &HPAGE_MASK );
lib/s390x/sie.h +2 -0
@@ -205,12 +205,14 @@ union {struct vm_uv {uint64_t vm_handle;uint64_t vcpu_handle; +uint64_t asce;void *conf_base_stor;void *conf_var_stor;void *cpu_stor; };struct vm_save_regs { +uint64_t asce;uint64_t grs[16 ];uint64_t fprs[16 ];uint32_t fpc;
lib/s390x/uv.c +33 -2
@@ -76,7 +76,8 @@ void uv_init(void)int cc;/* Let's not do this twice */ -assert (!initialized); +if (initialized) +return ;/* Query is done on initialization but let's check anyway */ assert (uvcb_qui.header.rc ==1 || uvcb_qui.header.rc ==0x100 );
@@ -90,6 +91,25 @@ void uv_init(void) initialized = true; } +/* +* Create a new ASCE for the UV config because they can't be shared +* for security reasons. We just simply copy the top most table into a +* fresh set of allocated pages and use those pages as the asce. +*/ +static uint64_t create_asce (void ) +{ +void *pgd_new, *pgd_old; +uint64_t asce =stctg (1 ); + + pgd_new =memalign_pages_flags (PAGE_SIZE ,PAGE_SIZE *4 ,0 ); + pgd_old = (void *)(asce &PAGE_MASK ); + +memcpy (pgd_new, pgd_old,PAGE_SIZE *4 ); + + asce =__pa (pgd_new) |ASCE_DT_REGION1 |REGION_TABLE_LENGTH |ASCE_P ; +return asce; +} +void uv_create_guest (struct vm *vm) {struct uv_cb_cgc uvcb_cgc = {
@@ -125,7 +145,8 @@ void uv_create_guest(struct vm *vm) vm->uv.cpu_stor =memalign_pages_flags (PAGE_SIZE , uvcb_qui.cpu_stor_len,0 ); uvcb_csc.stor_origin = (uint64_t )vm->uv.cpu_stor; - uvcb_cgc.guest_asce = (uint64_t )stctg (1 ); + uvcb_cgc.guest_asce =create_asce (); + vm->save_area.guest.asce = uvcb_cgc.guest_asce; uvcb_cgc.guest_sca = (uint64_t )vm->sca; cc =uv_call (0 , (uint64_t )&uvcb_cgc);
@@ -166,6 +187,16 @@ void uv_destroy_guest(struct vm *vm)assert (cc ==0 );free_pages (vm->uv.conf_base_stor);free_pages (vm->uv.conf_var_stor); + +free_pages ((void *)(vm->uv.asce &PAGE_MASK )); +memset (&vm->uv,0 ,sizeof (vm->uv)); + +/* Convert the sblk back to non-PV */ + vm->save_area.guest.asce =stctg (1 ); + vm->sblk->sdf =0 ; + vm->sblk->sidad =0 ; + vm->sblk->pv_handle_cpu =0 ; + vm->sblk->pv_handle_config =0 ; }int uv_unpack (struct vm *vm,uint64_t addr,uint64_t len,uint64_t tweak)
lib/s390x/uv.h +2 -3
@@ -28,9 +28,8 @@ static inline void uv_setup_asces(void)/* We need to have a valid primary ASCE to run guests. */ setup_vm (); -/* Set P bit in ASCE as it is required for PV guests */ - asce =stctg (1 ) |ASCE_P ; -lctlg (1 , asce); +/* Grab the ASCE which setup_vm() just set up */ + asce =stctg (1 );/* Copy ASCE into home space CR */ lctlg (13 , asce);
s390x/cpu.S +6 -0
@@ -76,6 +76,9 @@ sie64a: .endr stfpc SIE_SAVEAREA_HOST_FPC(%r3) + stctg %c1, %c1, SIE_SAVEAREA_HOST_ASCE(%r3) + lctlg %c1, %c1, SIE_SAVEAREA_GUEST_ASCE(%r3) + # Store scb and save_area pointer into stack frame stg %r2,__SF_SIE_CONTROL(%r15) # save control block pointer stg %r3,__SF_SIE_SAVEAREA(%r15) # save guest register save area
@@ -102,6 +105,9 @@ sie_exit: # Load guest register save area lg %r14,__SF_SIE_SAVEAREA(%r15) + # Restore the host asce + lctlg %c1, %c1, SIE_SAVEAREA_HOST_ASCE(%r14) + # Store guest's gprs, fprs and fpc stmg %r0,%r13,SIE_SAVEAREA_GUEST_GRS(%r14) # save guest gprs 0-13 .irp i, 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15
Signature · 1 line
2.34.1
Re: [kvm-unit-tests PATCH 5/6] lib: s390x: Use a new asce for each PV guest
Quoting Janosch Frank (2022-07-29 10:26:32)
Quoted 6 lines
Every PV guest needs its own ASCE so let's copy the topmost table designated by CR1 to create a new ASCE for the PV guest. Before and after SIE we now need to switch ASCEs to and from the PV guest / test ASCE. The SIE assembly function does that automatically.
Signed-off-by Janosch Frank <[email protected]>
with two tiny things for you to consider.
[...]
lib/s390x/uv.c void uv_init(void) · line 80
Quoted 11 lines
lib/s390x/uv.c +2
-1@@ -76,7 +76,8 @@ void uv_init(void)int cc;/* Let's not do this twice */ -assert (!initialized); +if (initialized) +return ;
Caught me a bit by surprise, maybe you want to point out this change in the commit message.
[...]
Quoted 6 lines
lib/s390x/uv.c +6
-0+/* +* Create a new ASCE for the UV config because they can't be shared +* for security reasons. We just simply copy the top most table into a +* fresh set of allocated pages and use those pages as the asce. +*/ +static uint64_t create_asce (void )
lib/s390x/uv.c void uv_init(void) · line 109
Quoted 10 lines
lib/s390x/uv.c +10
-0+{ +void *pgd_new, *pgd_old; +uint64_t asce =stctg (1 ); + + pgd_new =memalign_pages_flags (PAGE_SIZE ,PAGE_SIZE *4 ,0 ); + pgd_old = (void *)(asce &PAGE_MASK ); + +memcpy (pgd_new, pgd_old,PAGE_SIZE *4 ); + + asce =__pa (pgd_new) |ASCE_DT_REGION1 |REGION_TABLE_LENGTH |ASCE_P ;
I am wondering whether it might make sense to copy over the flags from cr1 so we don't have to worry about keeping them in sync?
But possibly it is sufficiently unlikely that these will ever change...
PATCH 6/6 lib: s390x: sie: Properly populate SCA
In reply to Janosch FrankCPU0 is the only cpu that's being used but we should still mark it as online and set the SDA in the SCA.
--- lib/s390x/sie.c | 8 ++++++++ lib/s390x/sie.h | 35 ++++++++++++++++++++++++++++++++++- 2 files changed, 42 insertions(+), 1 deletion(-)
lib/s390x/sie.c +8 -0
@@ -9,6 +9,7 @@*/ #include <asm/barrier.h> +#include <bitops.h> #include <libcflat.h> #include <sie.h> #include <asm/page.h>
@@ -71,6 +72,13 @@ void sie_guest_sca_create(struct vm *vm) vm->sblk->scaoh = ((uint64_t )vm->sca >>32 ); vm->sblk->scaol = (uint64_t )vm->sca & ~0x3fU ; vm->sblk->ecb2 |=ECB2_ESCA ; + +/* Enable SIGP sense running interpretation */ + vm->sblk->ecb |=ECB_SRSI ; + +/* We assume that cpu 0 is always part of the vm */ + vm->sca->mcn[0 ] =BIT (63 ); + vm->sca->cpu[0 ].sda = (uint64_t )vm->sblk; }/* Initializes the struct vm members like the SIE control block. */
lib/s390x/sie.h +34 -1
@@ -202,6 +202,39 @@ union {uint64_t pv_grregs[16 ];/* 0x0380 */ } __attribute__((packed)); +union esca_sigp_ctrl { +uint16_t value; +struct { +uint8_t c :1 ; +uint8_t reserved:7 ; +uint8_t scn; + }; +}; + +struct esca_entry { +union esca_sigp_ctrl sigp_ctrl; +uint16_t reserved1[3 ]; +uint64_t sda; +uint64_t reserved2[6 ]; +}; + +union ipte_control { +unsigned long val; +struct { +unsigned long k :1 ; +unsigned long kh :31 ; +unsigned long kg :32 ; + }; +}; + +struct esca_block { +union ipte_control ipte_control; +uint64_t reserved1[7 ]; +uint64_t mcn[4 ]; +uint64_t reserved2[20 ]; +struct esca_entry cpu[256 ]; +}; +struct vm_uv {uint64_t vm_handle;uint64_t vcpu_handle;
@@ -230,7 +263,7 @@ struct vm_save_area {struct vm {struct kvm_s390_sie_block *sblk;struct vm_save_area save_area; -void *sca;/* System Control Area */ +struct esca_block *sca;/* System Control Area */ uint8_t *crycb;/* Crypto Control Block */ struct vm_uv uv;/* PV UV information */ /* Ptr to first guest page */
Signature · 1 line
2.34.1
Re: [kvm-unit-tests PATCH 6/6] lib: s390x: sie: Properly populate SCA
Quoting Janosch Frank (2022-07-29 10:26:33)
Quoted 4 lines
CPU0 is the only cpu that's being used but we should still mark it as online and set the SDA in the SCA.
Signed-off-by Janosch Frank <[email protected]>