[PATCH bpf] selftests: bpf: add zero extend checks for ALU32 and/or/xor

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>
---
 .../selftests/bpf/verifier/basic_instr.c      | 39 +++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/tools/testing/selftests/bpf/verifier/basic_instr.c b/tools/testing/selftests/bpf/verifier/basic_instr.c
index ed91a7b9a456..4d844089938e 100644
--- a/tools/testing/selftests/bpf/verifier/basic_instr.c
+++ b/tools/testing/selftests/bpf/verifier/basic_instr.c

@@ -132,3 +132,42 @@
 	.prog_type = BPF_PROG_TYPE_SCHED_CLS,
 	.result = ACCEPT,
 },
+{
+	"and32 reg zero extend check",
+	.insns = {
+	BPF_MOV64_IMM(BPF_REG_0, -1),
+	BPF_MOV64_IMM(BPF_REG_2, -2),
+	BPF_ALU32_REG(BPF_AND, BPF_REG_0, BPF_REG_2),
+	BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+	BPF_EXIT_INSN(),
+	},
+	.prog_type = BPF_PROG_TYPE_SCHED_CLS,
+	.result = ACCEPT,
+	.retval = 0,
+},
+{
+	"or32 reg zero extend check",
+	.insns = {
+	BPF_MOV64_IMM(BPF_REG_0, -1),
+	BPF_MOV64_IMM(BPF_REG_2, -2),
+	BPF_ALU32_REG(BPF_OR, BPF_REG_0, BPF_REG_2),
+	BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+	BPF_EXIT_INSN(),
+	},
+	.prog_type = BPF_PROG_TYPE_SCHED_CLS,
+	.result = ACCEPT,
+	.retval = 0,
+},
+{
+	"xor32 reg zero extend check",
+	.insns = {
+	BPF_MOV64_IMM(BPF_REG_0, -1),
+	BPF_MOV64_IMM(BPF_REG_2, 0),
+	BPF_ALU32_REG(BPF_XOR, BPF_REG_0, BPF_REG_2),
+	BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+	BPF_EXIT_INSN(),
+	},
+	.prog_type = BPF_PROG_TYPE_SCHED_CLS,
+	.result = ACCEPT,
+	.retval = 0,
+},

-- 
2.20.1

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

On Wed, 22 May 2019 at 20:13, Y Song [off-list ref] wrote:

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

quoted

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

Thank you. Actually, it was intended for the bpf tree, as a test
follow up for this [1] fix.


Cheers,
Björn

[1] https://lore.kernel.org/bpf/CAJ+HfNifkxKz8df7gLBuqWA6+t6awrrRK6oW6m1nAYETJD+Vfg@mail.gmail.com/

quoted

---
 .../selftests/bpf/verifier/basic_instr.c      | 39 +++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/tools/testing/selftests/bpf/verifier/basic_instr.c b/tools/testing/selftests/bpf/verifier/basic_instr.c
index ed91a7b9a456..4d844089938e 100644
--- a/tools/testing/selftests/bpf/verifier/basic_instr.c
+++ b/tools/testing/selftests/bpf/verifier/basic_instr.c

@@ -132,3 +132,42 @@
        .prog_type = BPF_PROG_TYPE_SCHED_CLS,
        .result = ACCEPT,
 },
+{
+       "and32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, -2),
+       BPF_ALU32_REG(BPF_AND, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
+{
+       "or32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, -2),
+       BPF_ALU32_REG(BPF_OR, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
+{
+       "xor32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, 0),
+       BPF_ALU32_REG(BPF_XOR, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
--

2.20.1

On Wed, May 22, 2019 at 1:46 PM Björn Töpel [off-list ref] wrote:

On Wed, 22 May 2019 at 20:13, Y Song [off-list ref] wrote:

quoted

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

quoted

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

Thank you. Actually, it was intended for the bpf tree, as a test
follow up for this [1] fix.

Then maybe you want to add a Fixes tag and resubmit?

Cheers,
Björn

[1] https://lore.kernel.org/bpf/CAJ+HfNifkxKz8df7gLBuqWA6+t6awrrRK6oW6m1nAYETJD+Vfg@mail.gmail.com/

quoted

quoted

---
 .../selftests/bpf/verifier/basic_instr.c      | 39 +++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/tools/testing/selftests/bpf/verifier/basic_instr.c b/tools/testing/selftests/bpf/verifier/basic_instr.c
index ed91a7b9a456..4d844089938e 100644
--- a/tools/testing/selftests/bpf/verifier/basic_instr.c
+++ b/tools/testing/selftests/bpf/verifier/basic_instr.c

@@ -132,3 +132,42 @@
        .prog_type = BPF_PROG_TYPE_SCHED_CLS,
        .result = ACCEPT,
 },
+{
+       "and32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, -2),
+       BPF_ALU32_REG(BPF_AND, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
+{
+       "or32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, -2),
+       BPF_ALU32_REG(BPF_OR, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
+{
+       "xor32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, 0),
+       BPF_ALU32_REG(BPF_XOR, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
--

2.20.1

On Thu, 23 May 2019 at 08:39, Y Song [off-list ref] wrote:

On Wed, May 22, 2019 at 1:46 PM Björn Töpel [off-list ref] wrote:

quoted

On Wed, 22 May 2019 at 20:13, Y Song [off-list ref] wrote:

quoted

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

quoted

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

Thank you. Actually, it was intended for the bpf tree, as a test
follow up for this [1] fix.

Then maybe you want to add a Fixes tag and resubmit?

Hmm, I thought that adding tests were OK for non-next. Should the
Fixes: tag for the test reflex the corresponding fixed code (in this
case the RV JIT)?

quoted

Cheers,
Björn

[1] https://lore.kernel.org/bpf/CAJ+HfNifkxKz8df7gLBuqWA6+t6awrrRK6oW6m1nAYETJD+Vfg@mail.gmail.com/

quoted

quoted

---
 .../selftests/bpf/verifier/basic_instr.c      | 39 +++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/tools/testing/selftests/bpf/verifier/basic_instr.c b/tools/testing/selftests/bpf/verifier/basic_instr.c
index ed91a7b9a456..4d844089938e 100644
--- a/tools/testing/selftests/bpf/verifier/basic_instr.c
+++ b/tools/testing/selftests/bpf/verifier/basic_instr.c

@@ -132,3 +132,42 @@
        .prog_type = BPF_PROG_TYPE_SCHED_CLS,
        .result = ACCEPT,
 },
+{
+       "and32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, -2),
+       BPF_ALU32_REG(BPF_AND, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
+{
+       "or32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, -2),
+       BPF_ALU32_REG(BPF_OR, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
+{
+       "xor32 reg zero extend check",
+       .insns = {
+       BPF_MOV64_IMM(BPF_REG_0, -1),
+       BPF_MOV64_IMM(BPF_REG_2, 0),
+       BPF_ALU32_REG(BPF_XOR, BPF_REG_0, BPF_REG_2),
+       BPF_ALU64_IMM(BPF_RSH, BPF_REG_0, 32),
+       BPF_EXIT_INSN(),
+       },
+       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
+       .result = ACCEPT,
+       .retval = 0,
+},
--

2.20.1

On 05/23/2019 08:38 AM, Y Song wrote:

On Wed, May 22, 2019 at 1:46 PM Björn Töpel [off-list ref] wrote:

quoted

On Wed, 22 May 2019 at 20:13, Y Song [off-list ref] wrote:

quoted

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

quoted

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

Thank you. Actually, it was intended for the bpf tree, as a test
follow up for this [1] fix.

Then maybe you want to add a Fixes tag and resubmit?

Why would the test case need a fixes tag? It's common practice that we have
BPF fixes that we queue to bpf tree along with kselftest test cases related
to them. Therefore, applied as well, thanks for following up!

Björn, in my email from the fix, I mentioned we should have test snippets
ideally for all of the alu32 insns to not miss something falling through the
cracks when JITs get added or changed. If you have some cycles to add the
remaining missing ones, that would be much appreciated.

Thanks,
Daniel

On 23 May 2019, at 15:02, Daniel Borkmann [off-list ref] wrote:

On 05/23/2019 08:38 AM, Y Song wrote:

quoted

On Wed, May 22, 2019 at 1:46 PM Björn Töpel [off-list ref] wrote:

quoted

On Wed, 22 May 2019 at 20:13, Y Song [off-list ref] wrote:

quoted

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

quoted

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

Thank you. Actually, it was intended for the bpf tree, as a test
follow up for this [1] fix.

Then maybe you want to add a Fixes tag and resubmit?

Why would the test case need a fixes tag? It's common practice that we have
BPF fixes that we queue to bpf tree along with kselftest test cases related
to them. Therefore, applied as well, thanks for following up!

Björn, in my email from the fix, I mentioned we should have test snippets
ideally for all of the alu32 insns to not miss something falling through the
cracks when JITs get added or changed. If you have some cycles to add the
remaining missing ones, that would be much appreciated.

Björn,

  If you don’t have time, I can take this alu32 test case follow up as well.

Regards,
Jiong

Thanks,
Daniel

On Thu, 23 May 2019 at 16:31, Jiong Wang [off-list ref] wrote:

quoted

On 23 May 2019, at 15:02, Daniel Borkmann [off-list ref] wrote:

On 05/23/2019 08:38 AM, Y Song wrote:

quoted

On Wed, May 22, 2019 at 1:46 PM Björn Töpel [off-list ref] wrote:

quoted

On Wed, 22 May 2019 at 20:13, Y Song [off-list ref] wrote:

quoted

On Wed, May 22, 2019 at 2:25 AM Björn Töpel [off-list ref] wrote:

quoted

Add three tests to test_verifier/basic_instr that make sure that the
high 32-bits of the destination register is cleared after an ALU32
and/or/xor.

Signed-off-by: Björn Töpel <redacted>

I think the patch intends for bpf-next, right? The patch itself looks
good to me.
Acked-by: Yonghong Song <redacted>

Thank you. Actually, it was intended for the bpf tree, as a test
follow up for this [1] fix.

Then maybe you want to add a Fixes tag and resubmit?

Why would the test case need a fixes tag? It's common practice that we have
BPF fixes that we queue to bpf tree along with kselftest test cases related
to them. Therefore, applied as well, thanks for following up!

Björn, in my email from the fix, I mentioned we should have test snippets
ideally for all of the alu32 insns to not miss something falling through the
cracks when JITs get added or changed. If you have some cycles to add the
remaining missing ones, that would be much appreciated.

Björn,

  If you don’t have time, I can take this alu32 test case follow up as well.

Jiong, that would be great. Thank you. I'd guess it would take much
longer for me to do it (gmail.com time vs intel.com time ;-)).


Björn

Regards,
Jiong

quoted

Thanks,
Daniel