Skip to contentSkip to conversation

s390x: Add tests for execute-type instructions

PATCH Superseded Nina Schoetterl-Glausch 2 participants 4 messages Mar 10, 2023 – Mar 14, 2023

Subsystems (2)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

PATCH v5 1/1 s390x: Add tests for execute-type instructions

Test the instruction address used by targets of an execute instruction. When the target instruction calculates a relative address, the result is relative to the target instruction, not the execute instruction.

---

v4 -> v5:
* word align the execute-type instruction, preventing a specification exception if the address calculation is wrong, since LLGFRL requires word alignment
* change wording of comment

v3 -> v4:
* fix nits (thanks Janosch)
* pickup R-b (thanks Janosch)

v2 -> v3:
* add some comments (thanks Janosch)
* add two new tests (drop Nico's R-b)
* push prefix

v1 -> v2:
* add test to unittests.cfg and .gitlab-ci.yml
* pick up R-b (thanks Nico)

TCG does the address calculation relative to the execute instruction. Everything that has an operand that is relative to the instruction given by the immediate in the instruction and goes through in2_ri2 in TCG has this problem, because in2_ri2 does the calculation relative to pc_next which is the address of the EX(RL).
That should make fixing it easier tho.

Range-diff against v4:
1: f29ef634 ! 1: 57f8f256 s390x: Add tests for execute-type instructions

    @@ s390x/ex.c (new)
     +		"	.popsection\n"
     +
     +		"	llgfrl	%[target],0b\n"
    ++		//align (pad with nop), in case the wrong operand is used
    ++		"	.balignw 4,0x0707\n"
     +		"	exrl	0,0b\n"
     +		: [target] "=d" (target),
     +		  [value] "=d" (value)
    @@ s390x/ex.c (new)
     +		"	.popsection\n"
     +
     +		"	lrl	%[crl_word],0b\n"
    -+		//align (pad with nop), in case the wrong bad operand is used
    ++		//align (pad with nop), in case the wrong operand is used
     +		"	.balignw 4,0x0707\n"
     +		"	exrl	0,0b\n"
     +		"	ipm	%[program_mask]\n"
 s390x/Makefile      |   1 +
 s390x/ex.c          | 172 ++++++++++++++++++++++++++++++++++++++++++++
 s390x/unittests.cfg |   3 +
 .gitlab-ci.yml      |   1 +
 4 files changed, 177 insertions(+)
 create mode 100644 s390x/ex.c
s390x/Makefile +1 -0
index 97a61611..6cf8018b 100644
@@ -39,6 +39,7 @@ tests += $(TEST_DIR)/panic-loop-extint.elf
 tests += $(TEST_DIR)/panic-loop-pgm.elf
 tests += $(TEST_DIR)/migration-sck.elf
 tests += $(TEST_DIR)/exittime.elf
+tests += $(TEST_DIR)/ex.elf
 
 pv-tests += $(TEST_DIR)/pv-diags.elf
 
s390x/ex.c new file +172 -0

172 lines not on this page · raw message

s390x/unittests.cfg +3 -0
index d97eb5e9..b61faf07 100644
@@ -215,3 +215,6 @@ file = migration-skey.elf
 smp = 2
 groups = migration
 extra_params = -append '--parallel'
+
+[execute]
+file = ex.elf
.gitlab-ci.yml +1 -0
index ad7949c9..a999f64a 100644
@@ -275,6 +275,7 @@ s390x-kvm:
   - ACCEL=kvm ./run_tests.sh
       selftest-setup intercept emulator sieve sthyi diag10 diag308 pfmf
       cmm vector gs iep cpumodel diag288 stsi sclp-1g sclp-3g css skrf sie
+      execute
       | tee results.txt
   - grep -q PASS results.txt && ! grep -q FAIL results.txt
  only:

base-commit: e3c5c3ef2524c58023073c0fadde2e8ae3c04ec6

Signature · 1 line
2.39.1
#1 lore raw
Quoted 100 lines · On Fri, 10 Mar 2023 19:11:31 +0100 Nina Schoetterl-Glausch <[email protected]>…

Test the instruction address used by targets of an execute instruction. When the target instruction calculates a relative address, the result is relative to the target instruction, not the execute instruction.

---

v4 -> v5:
* word align the execute-type instruction, preventing a specification exception if the address calculation is wrong, since LLGFRL requires word alignment
* change wording of comment

v3 -> v4:
* fix nits (thanks Janosch)
* pickup R-b (thanks Janosch)

v2 -> v3:
* add some comments (thanks Janosch)
* add two new tests (drop Nico's R-b)
* push prefix

v1 -> v2:
* add test to unittests.cfg and .gitlab-ci.yml
* pick up R-b (thanks Nico)

TCG does the address calculation relative to the execute instruction. Everything that has an operand that is relative to the instruction given by the immediate in the instruction and goes through in2_ri2 in TCG has this problem, because in2_ri2 does the calculation relative to pc_next which is the address of the EX(RL).
That should make fixing it easier tho.

Range-diff against v4:
1: f29ef634 ! 1: 57f8f256 s390x: Add tests for execute-type instructions

    @@ s390x/ex.c (new)
     +		"	.popsection\n"
     +
     +		"	llgfrl	%[target],0b\n"
    ++		//align (pad with nop), in case the wrong operand is used
    ++		"	.balignw 4,0x0707\n"
     +		"	exrl	0,0b\n"
     +		: [target] "=d" (target),
     +		  [value] "=d" (value)
    @@ s390x/ex.c (new)
     +		"	.popsection\n"
     +
     +		"	lrl	%[crl_word],0b\n"
    -+		//align (pad with nop), in case the wrong bad operand is used
    ++		//align (pad with nop), in case the wrong operand is used
     +		"	.balignw 4,0x0707\n"
     +		"	exrl	0,0b\n"
     +		"	ipm	%[program_mask]\n"
 s390x/Makefile      |   1 +
 s390x/ex.c          | 172 ++++++++++++++++++++++++++++++++++++++++++++
 s390x/unittests.cfg |   3 +
 .gitlab-ci.yml      |   1 +
 4 files changed, 177 insertions(+)
 create mode 100644 s390x/ex.c
s390x/Makefile +1 -0
index 97a61611..6cf8018b 100644
@@ -39,6 +39,7 @@ tests += $(TEST_DIR)/panic-loop-extint.elf
 tests += $(TEST_DIR)/panic-loop-pgm.elf
 tests += $(TEST_DIR)/migration-sck.elf
 tests += $(TEST_DIR)/exittime.elf
+tests += $(TEST_DIR)/ex.elf
 
 pv-tests += $(TEST_DIR)/pv-diags.elf
 
s390x/ex.c new file +17 -0
new file mode 100644
index 00000000..f05f8f90
@@ -0,0 +1,172 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright IBM Corp. 2023
+ *
+ * Test EXECUTE (RELATIVE LONG).
+ * These instructions execute a target instruction. The target instruction is formed
+ * by reading an instruction from memory and optionally modifying some of its bits.
+ * The execution of the target instruction is the same as if it was executed
+ * normally as part of the instruction sequence, except for the instruction
+ * address and the instruction-length code.
+ */
+
+#include <libcflat.h>
+
+/*
+ * BRANCH AND SAVE, register register variant.
+ * Saves the next instruction address (address from PSW + length of instruction)

s390x/ex.c line 27

Quoted 10 lines
s390x/ex.c new file +10 -0
+ * to the first register. No branch is taken in this test, because 0 is
+ * specified as target.
+ * BASR does *not* perform a relative address calculation with an intermediate.
+ */
+static void test_basr(void)
+{
+	uint64_t ret_addr, after_ex;
+
+	report_prefix_push("BASR");
+	asm volatile ( ".pushsection .rodata\n"

you use .text.ex_bras in the next test, why not something like that here (and everywhere else) too?

Quoted 30 lines
s390x/ex.c new file +30 -0
+		"0:	basr	%[ret_addr],0\n"
+		"	.popsection\n"
+
+		"	larl	%[after_ex],1f\n"
+		"	exrl	0,0b\n"
+		"1:\n"
+		: [ret_addr] "=d" (ret_addr),
+		  [after_ex] "=d" (after_ex)
+	);
+
+	report(ret_addr == after_ex, "return address after EX");
+	report_prefix_pop();
+}
+
+/*
+ * BRANCH RELATIVE AND SAVE.
+ * According to PoP (Branch-Address Generation), the address calculated relative
+ * to the instruction address is relative to BRAS when it is the target of an
+ * execute-type instruction, not relative to the execute-type instruction.
+ */
+static void test_bras(void)
+{
+	uint64_t after_target, ret_addr, after_ex, branch_addr;
+
+	report_prefix_push("BRAS");
+	asm volatile ( ".pushsection .text.ex_bras, \"x\"\n"
+		"0:	bras	%[ret_addr],1f\n"
+		"	nopr	%%r7\n"
+		"1:	larl	%[branch_addr],0\n"
+		"	j	4f\n"

s390x/ex.c line 67

Quoted 10 lines
s390x/ex.c new file +10 -0
+		"	.popsection\n"
+
+		"	larl	%[after_target],1b\n"
+		"	larl	%[after_ex],3f\n"
+		"2:	exrl	0,0b\n"
+		"3:	larl	%[branch_addr],0\n"
+		"4:\n"
+
+		"	.if (1b - 0b) != (3b - 2b)\n"
+		"	.error	\"right and wrong target must have same offset\"\n"

please explain why briefly (i.e. if the wrong target is executed and the offset mismatches Bad Things™ happen)

Quoted 37 lines
s390x/ex.c new file +37 -0
+		"	.endif\n"
+		: [after_target] "=d" (after_target),
+		  [ret_addr] "=d" (ret_addr),
+		  [after_ex] "=d" (after_ex),
+		  [branch_addr] "=d" (branch_addr)
+	);
+
+	report(after_target == branch_addr, "address calculated relative to BRAS");
+	report(ret_addr == after_ex, "return address after EX");
+	report_prefix_pop();
+}
+
+/*
+ * LOAD ADDRESS RELATIVE LONG.
+ * If it is the target of an execute-type instruction, the address is relative
+ * to the LARL.
+ */
+static void test_larl(void)
+{
+	uint64_t target, addr;
+
+	report_prefix_push("LARL");
+	asm volatile ( ".pushsection .rodata\n"
+		"0:	larl	%[addr],0\n"
+		"	.popsection\n"
+
+		"	larl	%[target],0b\n"
+		"	exrl	0,0b\n"
+		: [target] "=d" (target),
+		  [addr] "=d" (addr)
+	);
+
+	report(target == addr, "address calculated relative to LARL");
+	report_prefix_pop();
+}
+
+/* LOAD LOGICAL RELATIVE LONG.

s390x/ex.c line 114

Quoted 10 lines
s390x/ex.c new file +10 -0
+ * If it is the target of an execute-type instruction, the address is relative
+ * to the LLGFRL.
+ */
+static void test_llgfrl(void)
+{
+	uint64_t target, value;
+
+	report_prefix_push("LLGFRL");
+	asm volatile ( ".pushsection .rodata\n"
+		"	.balign	4\n"

explain the alignment (like you did in the test below)

Quoted 83 lines
s390x/ex.c new file +58 -0
+		"0:	llgfrl	%[value],0\n"
+		"	.popsection\n"
+
+		"	llgfrl	%[target],0b\n"
+		//align (pad with nop), in case the wrong operand is used
+		"	.balignw 4,0x0707\n"
+		"	exrl	0,0b\n"
+		: [target] "=d" (target),
+		  [value] "=d" (value)
+	);
+
+	report(target == value, "loaded correct value");
+	report_prefix_pop();
+}
+
+/*
+ * COMPARE RELATIVE LONG
+ * If it is the target of an execute-type instruction, the address is relative
+ * to the CRL.
+ */
+static void test_crl(void)
+{
+	uint32_t program_mask, cc, crl_word;
+
+	report_prefix_push("CRL");
+	asm volatile ( ".pushsection .rodata\n"
+		 //operand of crl must be word aligned
+		 "	.balign	4\n"
+		"0:	crl	%[crl_word],0\n"
+		"	.popsection\n"
+
+		"	lrl	%[crl_word],0b\n"
+		//align (pad with nop), in case the wrong operand is used
+		"	.balignw 4,0x0707\n"
+		"	exrl	0,0b\n"
+		"	ipm	%[program_mask]\n"
+		: [program_mask] "=d" (program_mask),
+		  [crl_word] "=d" (crl_word)
+		:: "cc"
+	);
+
+	cc = program_mask >> 28;
+	report(!cc, "operand compared to is relative to CRL");
+	report_prefix_pop();
+}
+
+int main(int argc, char **argv)
+{
+	report_prefix_push("ex");
+	test_basr();
+	test_bras();
+	test_larl();
+	test_llgfrl();
+	test_crl();
+	report_prefix_pop();
+
+	return report_summary();
+}
s390x/unittests.cfg +3 -0
index d97eb5e9..b61faf07 100644
@@ -215,3 +215,6 @@ file = migration-skey.elf
 smp = 2
 groups = migration
 extra_params = -append '--parallel'
+
+[execute]
+file = ex.elf
.gitlab-ci.yml +1 -0
index ad7949c9..a999f64a 100644
@@ -275,6 +275,7 @@ s390x-kvm:
   - ACCEL=kvm ./run_tests.sh
       selftest-setup intercept emulator sieve sthyi diag10 diag308 pfmf
       cmm vector gs iep cpumodel diag288 stsi sclp-1g sclp-3g css skrf sie
+      execute
       | tee results.txt
   - grep -q PASS results.txt && ! grep -q FAIL results.txt
  only:

base-commit: e3c5c3ef2524c58023073c0fadde2e8ae3c04ec6

#2 lore raw
Quoted 108 lines · Claudio Imbrenda
Quoted 105 lines · On Fri, 10 Mar 2023 19:11:31 +0100 Nina Schoetterl-Glausch <[email protected]>…

Test the instruction address used by targets of an execute instruction. When the target instruction calculates a relative address, the result is relative to the target instruction, not the execute instruction.

---

v4 -> v5:
* word align the execute-type instruction, preventing a specification exception if the address calculation is wrong, since LLGFRL requires word alignment
* change wording of comment

v3 -> v4:
* fix nits (thanks Janosch)
* pickup R-b (thanks Janosch)

v2 -> v3:
* add some comments (thanks Janosch)
* add two new tests (drop Nico's R-b)
* push prefix

v1 -> v2:
* add test to unittests.cfg and .gitlab-ci.yml
* pick up R-b (thanks Nico)

TCG does the address calculation relative to the execute instruction. Everything that has an operand that is relative to the instruction given by the immediate in the instruction and goes through in2_ri2 in TCG has this problem, because in2_ri2 does the calculation relative to pc_next which is the address of the EX(RL).
That should make fixing it easier tho.

Range-diff against v4:
1: f29ef634 ! 1: 57f8f256 s390x: Add tests for execute-type instructions

    @@ s390x/ex.c (new)
     +		"	.popsection\n"
     +
     +		"	llgfrl	%[target],0b\n"
    ++		//align (pad with nop), in case the wrong operand is used
    ++		"	.balignw 4,0x0707\n"
     +		"	exrl	0,0b\n"
     +		: [target] "=d" (target),
     +		  [value] "=d" (value)
    @@ s390x/ex.c (new)
     +		"	.popsection\n"
     +
     +		"	lrl	%[crl_word],0b\n"
    -+		//align (pad with nop), in case the wrong bad operand is used
    ++		//align (pad with nop), in case the wrong operand is used
     +		"	.balignw 4,0x0707\n"
     +		"	exrl	0,0b\n"
     +		"	ipm	%[program_mask]\n"
 s390x/Makefile      |   1 +
 s390x/ex.c          | 172 ++++++++++++++++++++++++++++++++++++++++++++
 s390x/unittests.cfg |   3 +
 .gitlab-ci.yml      |   1 +
 4 files changed, 177 insertions(+)
 create mode 100644 s390x/ex.c
s390x/Makefile +1 -0
index 97a61611..6cf8018b 100644
@@ -39,6 +39,7 @@ tests += $(TEST_DIR)/panic-loop-extint.elf
 tests += $(TEST_DIR)/panic-loop-pgm.elf
 tests += $(TEST_DIR)/migration-sck.elf
 tests += $(TEST_DIR)/exittime.elf
+tests += $(TEST_DIR)/ex.elf
 
 pv-tests += $(TEST_DIR)/pv-diags.elf
 
s390x/ex.c new file +22 -0
new file mode 100644
index 00000000..f05f8f90
@@ -0,0 +1,172 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright IBM Corp. 2023
+ *
+ * Test EXECUTE (RELATIVE LONG).
+ * These instructions execute a target instruction. The target instruction is formed
+ * by reading an instruction from memory and optionally modifying some of its bits.
+ * The execution of the target instruction is the same as if it was executed
+ * normally as part of the instruction sequence, except for the instruction
+ * address and the instruction-length code.
+ */
+
+#include <libcflat.h>
+
+/*
+ * BRANCH AND SAVE, register register variant.
+ * Saves the next instruction address (address from PSW + length of instruction)
+ * to the first register. No branch is taken in this test, because 0 is
+ * specified as target.
+ * BASR does *not* perform a relative address calculation with an intermediate.
+ */
+static void test_basr(void)
Quoted 8 lines

s390x/ex.c line 27

Quoted 5 lines
s390x/ex.c new file +5 -0
+{
+	uint64_t ret_addr, after_ex;
+
+	report_prefix_push("BASR");
+	asm volatile ( ".pushsection .rodata\n"

you use .text.ex_bras in the next test, why not something like that here (and everywhere else) too?

In the test below we branch to the code in .text.ex_bras.
In all other tests the instruction in .rodata is just an operand of the execute instruction, and it doesn't get modified.
As for the bras test having a suffix, I guess it's pretty arbitrary, but since it's a handful of instructions instead of just one, it felt substantial enough to warrant one.

Quoted 28 lines
Quoted 28 lines
s390x/ex.c new file +27 -0
+		"0:	basr	%[ret_addr],0\n"
+		"	.popsection\n"
+
+		"	larl	%[after_ex],1f\n"
+		"	exrl	0,0b\n"
+		"1:\n"
+		: [ret_addr] "=d" (ret_addr),
+		  [after_ex] "=d" (after_ex)
+	);
+
+	report(ret_addr == after_ex, "return address after EX");
+	report_prefix_pop();
+}
+
+/*
+ * BRANCH RELATIVE AND SAVE.
+ * According to PoP (Branch-Address Generation), the address calculated relative
+ * to the instruction address is relative to BRAS when it is the target of an
+ * execute-type instruction, not relative to the execute-type instruction.
+ */
+static void test_bras(void)
+{
+	uint64_t after_target, ret_addr, after_ex, branch_addr;
+
+	report_prefix_push("BRAS");
+	asm volatile ( ".pushsection .text.ex_bras, \"x\"\n"
+		"0:	bras	%[ret_addr],1f\n"
Quoted 8 lines

s390x/ex.c line 62

Quoted 8 lines
s390x/ex.c new file +8 -0
+		"	nopr	%%r7\n"
+		"1:	larl	%[branch_addr],0\n"
+		"	j	4f\n"
+		"	.popsection\n"
+
+		"	larl	%[after_target],1b\n"
+		"	larl	%[after_ex],3f\n"
+		"2:	exrl	0,0b\n"

/*
* In case the address calculation is correct, we jump by the relative offset 1b-0b from 0b to 1b.
* In case the address calculation is relative to the exrl (i.e. a test failure),
* put a valid instruction at the same relative offset from the exrl, so the test continues in a
* controlled manner.
*/

Quoted 8 lines

s390x/ex.c line 67

Quoted 5 lines
s390x/ex.c new file +5 -0
+		"3:	larl	%[branch_addr],0\n"
+		"4:\n"
+
+		"	.if (1b - 0b) != (3b - 2b)\n"
+		"	.error	\"right and wrong target must have same offset\"\n"

please explain why briefly (i.e. if the wrong target is executed and the offset mismatches Bad Things™ happen)

Ok, see above.

[...]

#3 lore raw

On Mon, 13 Mar 2023 23:45:33 +0100
Nina Schoetterl-Glausch <[email protected]> wrote:

[...]

Quoted 17 lines
Quoted 8 lines
Quoted 8 lines
s390x/ex.c new file +8 -0
+/*
+ * BRANCH AND SAVE, register register variant.
+ * Saves the next instruction address (address from PSW + length of instruction)
+ * to the first register. No branch is taken in this test, because 0 is
+ * specified as target.
+ * BASR does *not* perform a relative address calculation with an intermediate.
+ */
+static void test_basr(void)
Quoted 8 lines

s390x/ex.c line 27

Quoted 5 lines
s390x/ex.c new file +5 -0
+{
+	uint64_t ret_addr, after_ex;
+
+	report_prefix_push("BASR");
+	asm volatile ( ".pushsection .rodata\n"  

you use .text.ex_bras in the next test, why not something like that here (and everywhere else) too?

Quoted 6 lines

In the test below we branch to the code in .text.ex_bras.
In all other tests the instruction in .rodata is just an operand of the execute instruction, and it doesn't get modified.
As for the bras test having a suffix, I guess it's pretty arbitrary, but since it's a handful of instructions instead of just one, it felt substantial enough to warrant one.

we discussed this offline :)

Quoted 34 lines
Quoted 34 lines
Quoted 34 lines
s390x/ex.c new file +33 -0
+		"0:	basr	%[ret_addr],0\n"
+		"	.popsection\n"
+
+		"	larl	%[after_ex],1f\n"
+		"	exrl	0,0b\n"
+		"1:\n"
+		: [ret_addr] "=d" (ret_addr),
+		  [after_ex] "=d" (after_ex)
+	);
+
+	report(ret_addr == after_ex, "return address after EX");
+	report_prefix_pop();
+}
+
+/*
+ * BRANCH RELATIVE AND SAVE.
+ * According to PoP (Branch-Address Generation), the address calculated relative
+ * to the instruction address is relative to BRAS when it is the target of an
+ * execute-type instruction, not relative to the execute-type instruction.
+ */
+static void test_bras(void)
+{
+	uint64_t after_target, ret_addr, after_ex, branch_addr;
+
+	report_prefix_push("BRAS");
+	asm volatile ( ".pushsection .text.ex_bras, \"x\"\n"
+		"0:	bras	%[ret_addr],1f\n"
+		"	nopr	%%r7\n"
+		"1:	larl	%[branch_addr],0\n"
+		"	j	4f\n"
+		"	.popsection\n"
+
+		"	larl	%[after_target],1b\n"
Quoted 8 lines
Quoted 2 lines

s390x/ex.c line 62

Quoted 2 lines
s390x/ex.c new file +2 -0
+		"	larl	%[after_ex],3f\n"
+		"2:	exrl	0,0b\n"  

/*
* In case the address calculation is correct, we jump by the relative offset 1b-0b from 0b to 1b.
* In case the address calculation is relative to the exrl (i.e. a test failure),
* put a valid instruction at the same relative offset from the exrl, so the test continues in a
* controlled manner.
*/

looks good

Quoted 14 lines
Quoted 8 lines

s390x/ex.c line 67

Quoted 5 lines
s390x/ex.c new file +5 -0
+		"3:	larl	%[branch_addr],0\n"
+		"4:\n"
+
+		"	.if (1b - 0b) != (3b - 2b)\n"
+		"	.error	\"right and wrong target must have same offset\"\n"  

please explain why briefly (i.e. if the wrong target is executed and the offset mismatches Bad Things™ happen)

Ok, see above.

[...]