Skip to contentSkip to conversation

TPM-TIS bios-tables-test

RFC Superseded Eric Auger 3 participants 27 messages Jun 1, 2020 – Jun 9, 2020

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Eric Auger

author
#0 lore raw

cover TPM-TIS bios-tables-test

This series adds the test of the TPM2 and DSDT tables when the TPM-TIS device gets instantiated.

The series also allows to validate changes in the TPM2 table generation done in "[PATCH v3 0/4] vTPM/aarch64 ACPI support", sent separately.

This depends on Stefan's "acpi: tpm: Do not build TCPA table for TPM 2".

This is in RFC state because I am not very familiar with the table generation process docommented in bios-tables-test.c.

I don't know how to get rid of the checkpatch errors:
"ERROR: Does not appear to be a unified-diff format patch"

Surprisingly, checkpatch argues when void tables are added in the same time as bios-tables-test-allowed-diff.h changes.

Most importantly, to be able to reuse tpm-emu code, I was forced to remove the TPM2_ST_NO_SESSIONS assert. I would be grateful if I could get some advises on the best way to address the issue.

Best Regards

Eric

This series can be found at:
https://github.com/eauger/qemu/tree/v5.0-tpm-acpi-tests-v1

Eric Auger (6):
test/tpm-emu: include sockets and channel headers in tpm-emu header tests/acpi: Add void tables for Q35/TPM-TIS bios-tables-test tests/acpi: Ignore TPM2.tis and DSDT.tis
tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS
bios-tables-test: Add Q35/TPM-TIS test
bios-tables-test: Generate reference tables for Q35/TPM-TIS

 tests/qtest/tpm-emu.h          |   3 ++
 tests/qtest/bios-tables-test.c |  60 +++++++++++++++++++++++++++++++++
 tests/qtest/tpm-emu.c          |   1 -
 tests/data/acpi/q35/DSDT.tis   | Bin 0 -> 8468 bytes
 tests/data/acpi/q35/TPM2.tis   | Bin 0 -> 76 bytes
 tests/qtest/Makefile.include   |   1 +
 6 files changed, 64 insertions(+), 1 deletion(-)
 create mode 100644 tests/data/acpi/q35/DSDT.tis
 create mode 100644 tests/data/acpi/q35/TPM2.tis
Signature · 1 line
2.20.1
author
#1 lore raw

RFC 1/6 test/tpm-emu: include sockets and channel headers in tpm-emu header

Include sockets and channel headers to that the header is self-contained.

---
 tests/qtest/tpm-emu.h | 3 +++
 1 file changed, 3 insertions(+)
tests/qtest/tpm-emu.h +3 -0
index a4f1d64226..73f3bed0c4 100644
@@ -16,6 +16,9 @@
 #define TPM_RC_FAILURE 0x101
 #define TPM2_ST_NO_SESSIONS 0x8001
 
+#include "qemu/sockets.h"
+#include "io/channel.h"
+
 struct tpm_hdr {
     uint16_t tag;
     uint32_t len;
Signature · 1 line
2.20.1
Reviewed-by on v1 1/6
#2 lore raw

Re: [RFC 1/6] test/tpm-emu: include sockets and channel headers in tpm-emu header

Quoted 4 lines · Eric Auger

Include sockets and channel headers to that the header is self-contained.

Quoted 18 lines
---
  tests/qtest/tpm-emu.h | 3 +++
  1 file changed, 3 insertions(+)
tests/qtest/tpm-emu.h +3 -0
index a4f1d64226..73f3bed0c4 100644
@@ -16,6 +16,9 @@
  #define TPM_RC_FAILURE 0x101
  #define TPM2_ST_NO_SESSIONS 0x8001
 
+#include "qemu/sockets.h"
+#include "io/channel.h"
+
  struct tpm_hdr {
      uint16_t tag;
      uint32_t len;
author
#3 lore raw

RFC 2/6 tests/acpi: Add void tables for Q35/TPM-TIS bios-tables-test

In reply to Eric Auger

Add placeholders for TPM and DSDT reference tables for Q35 TPM-TIS tests.

---
 tests/data/acpi/q35/DSDT.tis | Bin
 tests/data/acpi/q35/TPM2.tis | Bin
 2 files changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 tests/data/acpi/q35/DSDT.tis
 create mode 100644 tests/data/acpi/q35/TPM2.tis
tests/data/acpi/q35/DSDT.tis new file +0 -0
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
tests/data/acpi/q35/TPM2.tis new file +0 -0
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
Signature · 1 line
2.20.1
Reviewed-by on v1 2/6
#4 lore raw

Re: [RFC 2/6] tests/acpi: Add void tables for Q35/TPM-TIS bios-tables-test

Quoted 4 lines · Eric Auger

Add placeholders for TPM and DSDT reference tables for Q35 TPM-TIS tests.

Quoted 13 lines
---
  tests/data/acpi/q35/DSDT.tis | Bin
  tests/data/acpi/q35/TPM2.tis | Bin
  2 files changed, 0 insertions(+), 0 deletions(-)

create mode 100644 tests/data/acpi/q35/DSDT.tis
create mode 100644 tests/data/acpi/q35/TPM2.tis

tests/data/acpi/q35/DSDT.tis new file +0 -0
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
tests/data/acpi/q35/TPM2.tis new file +0 -0
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
Reviewed-by on v1 2/6
#5 lore raw

Re: [RFC 2/6] tests/acpi: Add void tables for Q35/TPM-TIS bios-tables-test

In reply to Eric Auger
Quoted 4 lines · On Mon, 1 Jun 2020 12:21:09 +0200 Eric Auger <[email protected]> wrote:

Add placeholders for TPM and DSDT reference tables for Q35 TPM-TIS tests.

Quoted 13 lines
---
 tests/data/acpi/q35/DSDT.tis | Bin
 tests/data/acpi/q35/TPM2.tis | Bin
 2 files changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 tests/data/acpi/q35/DSDT.tis
 create mode 100644 tests/data/acpi/q35/TPM2.tis
tests/data/acpi/q35/DSDT.tis new file +0 -0
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
tests/data/acpi/q35/TPM2.tis new file +0 -0
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
author
#6 lore raw

RFC 3/6 tests/acpi: Ignore TPM2.tis and DSDT.tis

In reply to Eric Auger

Subsequent patches will alter the content of TPM2.tis and DSDT.tis so let's ignore them until the references are generated.

---
 tests/qtest/bios-tables-test-allowed-diff.h | 2 ++
 1 file changed, 2 insertions(+)
tests/qtest/bios-tables-test-allowed-diff.h +2 -0
index dfb8523c8b..a2a45d1d31 100644
@@ -1 +1,3 @@
 /* List of comma-separated changed AML files to ignore */
+"tests/data/acpi/q35/DSDT.tis",
+"tests/data/acpi/q35/TPM2.tis",
Signature · 1 line
2.20.1
Reviewed-by on v1 3/6
#7 lore raw

Re: [RFC 3/6] tests/acpi: Ignore TPM2.tis and DSDT.tis

Quoted 5 lines · Eric Auger

Subsequent patches will alter the content of TPM2.tis and DSDT.tis so let's ignore them until the references are generated.

Quoted 12 lines
---
  tests/qtest/bios-tables-test-allowed-diff.h | 2 ++
  1 file changed, 2 insertions(+)
tests/qtest/bios-tables-test-allowed-diff.h +2 -0
index dfb8523c8b..a2a45d1d31 100644
@@ -1 +1,3 @@
  /* List of comma-separated changed AML files to ignore */
+"tests/data/acpi/q35/DSDT.tis",
+"tests/data/acpi/q35/TPM2.tis",
Reviewed-by on v1 3/6
#8 lore raw

Re: [RFC 3/6] tests/acpi: Ignore TPM2.tis and DSDT.tis

In reply to Eric Auger
Quoted 5 lines · On Mon, 1 Jun 2020 12:21:10 +0200 Eric Auger <[email protected]> wrote:

Subsequent patches will alter the content of TPM2.tis and DSDT.tis so let's ignore them until the references are generated.

Quoted 12 lines
---
 tests/qtest/bios-tables-test-allowed-diff.h | 2 ++
 1 file changed, 2 insertions(+)
tests/qtest/bios-tables-test-allowed-diff.h +2 -0
index dfb8523c8b..a2a45d1d31 100644
@@ -1 +1,3 @@
 /* List of comma-separated changed AML files to ignore */
+"tests/data/acpi/q35/DSDT.tis",
+"tests/data/acpi/q35/TPM2.tis",
author
#9 lore raw

RFC 4/6 tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

In reply to Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
 tests/qtest/tpm-emu.c | 1 -
 1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -1
index c43ac4aef8..298d0eec74 100644
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
         s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
         s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
         g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);
 
         s->tpm_msg = g_realloc(s->tpm_msg, s->tpm_msg->len);
         qio_channel_read(ioc, (char *)&s->tpm_msg->code,
Signature · 1 line
2.20.1
#10 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
  tests/qtest/tpm-emu.c | 1 -
  1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
          s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
          s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
          g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

Quoted 3 lines
tests/qtest/tpm-emu.c +0 -0
          s->tpm_msg = g_realloc(s->tpm_msg, s->tpm_msg->len);
          qio_channel_read(ioc, (char *)&s->tpm_msg->code,
#11 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Quoted 14 lines · Stefan Berger
Quoted 13 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
  tests/qtest/tpm-emu.c | 1 -

  1 file changed, 1 deletion(-)
Quoted 10 lines

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 9 lines
tests/qtest/tpm-emu.c +0 -1
index c43ac4aef8..298d0eec74 100644
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
           s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
           s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
           g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

In hw/tpm/tpm_util.c tpm_util_test_tpmdev() we have a probing function that tries to determine whether the attached TPM is a TPM 1.2 or TPM 2. It sends a TPM 2 command first and I would say this is the only command we should see in this (fake) TPM emulation here. We respond with a TPM 2 command header [indicating failure], but the probing function should then not proceed to probe with the TPM 1.2 command.

Quoted 4 lines
Quoted 2 lines
tests/qtest/tpm-emu.c +0 -0
             s->tpm_msg = g_realloc(s->tpm_msg, s->tpm_msg->len);
           qio_channel_read(ioc, (char *)&s->tpm_msg->code,
author
#12 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

In reply to Stefan Berger

Hi Stefan,

Quoted 14 lines · Stefan Berger
Quoted 13 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
  tests/qtest/tpm-emu.c | 1 -

  1 file changed, 1 deletion(-)
Quoted 15 lines

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 9 lines
tests/qtest/tpm-emu.c +0 -1
index c43ac4aef8..298d0eec74 100644
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
           s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
           s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
           g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Thanks

Eric

Quoted 5 lines
Quoted 2 lines
tests/qtest/tpm-emu.c +0 -0
             s->tpm_msg = g_realloc(s->tpm_msg, s->tpm_msg->len);
           qio_channel_read(ioc, (char *)&s->tpm_msg->code,
#13 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Quoted 32 lines · Auger Eric

Hi Stefan,

Quoted 14 lines · Stefan Berger
Quoted 13 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
   tests/qtest/tpm-emu.c | 1 -
   1 file changed, 1 deletion(-)
Quoted 15 lines

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 9 lines
tests/qtest/tpm-emu.c +0 -1
index c43ac4aef8..298d0eec74 100644
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
           s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
           s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
           g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

   Stefan

author
#14 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Stefan,

Quoted 33 lines · Stefan Berger
Quoted 32 lines · Auger Eric

Hi Stefan,

Quoted 29 lines · Stefan Berger
Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
   tests/qtest/tpm-emu.c | 1 -

   1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
            s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
            s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
            g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

Quoted 3 lines
Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

From what I understand

QTEST_QEMU_BINARY=x86_64-softmmu/qemu-system-x86_64 tests/qtest/bios-tables-test

launches

x86_64-softmmu/qemu-system-x86_64 \
-qtest unix:/tmp/qtest-20181.sock \
-qtest-log /dev/null -chardev socket,path=/tmp/qtest-20181.qmp,id=char0 -mon chardev=char0,mode=control \
-display none -machine q35,kernel-irqchip=off \
-accel kvm -accel tcg \
-net none -display none \
-chardev socket,id=chr,path=/tmp/qemu-test_acpi_q35_tcg_tpm2.RJJCL0/sock -tpmdev emulator,id=dev,chardev=chr \
-device tpm-tis,tpmdev=dev \
-drive id=hd0,if=none,file=tests/acpi-test-disk-RAnagW,format=raw -device ide-hd,drive=hd0 -accel qtest

acpi-test-disk-RAnagW consists in an x86 boot sector code, created by boot_sector_init (x86_boot_sector)

Thanks

Eric

Quoted 5 lines

   Stefan

author
#15 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

In reply to Stefan Berger

Hi Stefan,

Quoted 33 lines · Stefan Berger
Quoted 32 lines · Auger Eric

Hi Stefan,

Quoted 29 lines · Stefan Berger
Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
   tests/qtest/tpm-emu.c | 1 -

   1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
            s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
            s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
            g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

Quoted 3 lines
Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

So it looks SeaBIOS is in use (bios-256k.bin loaded).

I can see MMIO accesses to the TPM and the following commands are observable:
tpm_emu_tpm_thread code=0x181 tag=0x8001 len=0xa tpm_emu_tpm_thread code=0x144 tag=0x8001 len=0xc tpm_emu_tpm_thread code=0x121 tag=0x8002 len=0x20
This last one causes the assert (TPM2_CC_HierarchyControl)

I checked in Seabios and effectively tpm20_hierarchycontrol() tags the TPM2_CC_HierarchyControl command with TPM2_ST_SESSIONS

Due to our emulation, maybe tpm_set_failure() gets called, inducing tpm20_hierarchycontrol() call.

That being said, what do you recommend? Remove the assert, improve the emulation, other?

Thank you in advance

Best Regards

Eric

Quoted 5 lines

   Stefan

#16 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Quoted 47 lines · Auger Eric

Hi Stefan,

Quoted 33 lines · Stefan Berger
Quoted 32 lines · Auger Eric

Hi Stefan,

Quoted 29 lines · Stefan Berger
Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
    tests/qtest/tpm-emu.c | 1 -
    1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
            s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
            s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
            g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you are seeing there.

Quoted 2 lines
Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

So it looks SeaBIOS is in use (bios-256k.bin loaded).

I can see MMIO accesses to the TPM and the following commands are observable:
tpm_emu_tpm_thread code=0x181 tag=0x8001 len=0xa tpm_emu_tpm_thread code=0x144 tag=0x8001 len=0xc tpm_emu_tpm_thread code=0x121 tag=0x8002 len=0x20
This last one causes the assert (TPM2_CC_HierarchyControl)

Quoted 8 lines

I checked in Seabios and effectively tpm20_hierarchycontrol() tags the TPM2_CC_HierarchyControl command with TPM2_ST_SESSIONS

Due to our emulation, maybe tpm_set_failure() gets called, inducing tpm20_hierarchycontrol() call.

That being said, what do you recommend? Remove the assert, improve the emulation, other?

So this is an ACPI test. What role does the firmware play for success of the test? If the test relies on the firmware showing some sort of expected result, then I would recommend only running this test with an attached swtpm, like we run some other tests. If we don't need the firmware to succeed then I would just get rid of the assert. Probably no other test we have implemented so far was running the firmware...

   Stefan

Quoted 11 lines

Thank you in advance

Best Regards

Eric

Quoted 4 lines

   Stefan

author
#17 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Hi Stefan,

Quoted 59 lines · Stefan Berger
Quoted 49 lines · Auger Eric

Hi Stefan,

Quoted 35 lines · Stefan Berger
Quoted 34 lines · Auger Eric

Hi Stefan,

Quoted 31 lines · Stefan Berger
Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
    tests/qtest/tpm-emu.c | 1 -

    1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
             s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
             s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
             g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);

You should not have to remove this. The tests are skipped if swtpm does
not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you
are seeing there.

Quoted 2 lines
Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

So it looks SeaBIOS is in use (bios-256k.bin loaded).

I can see MMIO accesses to the TPM and the following commands are observable:
tpm_emu_tpm_thread code=0x181 tag=0x8001 len=0xa tpm_emu_tpm_thread code=0x144 tag=0x8001 len=0xc tpm_emu_tpm_thread code=0x121 tag=0x8002 len=0x20
This last one causes the assert (TPM2_CC_HierarchyControl)

Quoted 8 lines

I checked in Seabios and effectively tpm20_hierarchycontrol() tags the TPM2_CC_HierarchyControl command with TPM2_ST_SESSIONS

Due to our emulation, maybe tpm_set_failure() gets called, inducing tpm20_hierarchycontrol() call.

That being said, what do you recommend? Remove the assert, improve the emulation, other?

Quoted 6 lines

So this is an ACPI test. What role does the firmware play for success of the test? If the test relies on the firmware showing some sort of expected result, then I would recommend only running this test with an attached swtpm, like we run some other tests. If we don't need the firmware to succeed then I would just get rid of the assert. Probably no other test we have implemented so far was running the firmware...

FWIU The goal of this test is to compare the acpi tables generated by qemu against reference ones. I dont think we expect from the FW any specific result but I would prefer Igor or Michael to confirm.

In that case, removing the assert() allows to compare the specific DSDT and TPM2 tables and that's our expectation here I think.

Thanks

Eric

Quoted 18 lines

   Stefan

Quoted 11 lines

Thank you in advance

Best Regards

Eric

Quoted 4 lines

    Stefan

#18 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Quoted 68 lines · On Fri, 5 Jun 2020 17:47:08 +0200 Auger Eric <[email protected]> wrote:

Hi Stefan,

Quoted 59 lines · Stefan Berger
Quoted 49 lines · Auger Eric

Hi Stefan,

Quoted 35 lines · Stefan Berger
Quoted 34 lines · Auger Eric

Hi Stefan,

Quoted 31 lines · Stefan Berger
Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
    tests/qtest/tpm-emu.c | 1 -

    1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
             s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
             s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
             g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);  

You should not have to remove this. The tests are skipped if swtpm does
not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you
are seeing there.

Quoted 2 lines
Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

So it looks SeaBIOS is in use (bios-256k.bin loaded).

I can see MMIO accesses to the TPM and the following commands are observable:
tpm_emu_tpm_thread code=0x181 tag=0x8001 len=0xa tpm_emu_tpm_thread code=0x144 tag=0x8001 len=0xc tpm_emu_tpm_thread code=0x121 tag=0x8002 len=0x20
This last one causes the assert (TPM2_CC_HierarchyControl)

Quoted 8 lines

I checked in Seabios and effectively tpm20_hierarchycontrol() tags the TPM2_CC_HierarchyControl command with TPM2_ST_SESSIONS

Due to our emulation, maybe tpm_set_failure() gets called, inducing tpm20_hierarchycontrol() call.

That being said, what do you recommend? Remove the assert, improve the emulation, other?

Quoted 6 lines

So this is an ACPI test. What role does the firmware play for success of the test? If the test relies on the firmware showing some sort of expected result, then I would recommend only running this test with an attached swtpm, like we run some other tests. If we don't need the firmware to succeed then I would just get rid of the assert. Probably no other test we have implemented so far was running the firmware...

Quoted 3 lines

FWIU The goal of this test is to compare the acpi tables generated by qemu against reference ones. I dont think we expect from the FW any specific result but I would prefer Igor or Michael to confirm.

Firmware is needed to fetch tables from QEMU and place them in guest RAM, it will also patch cross table pointers accordingly.

So bios-tables-test checks both QEMU and firmware at the same time, and reference tables are in form guest OS will see them.

Quoted 25 lines

In that case, removing the assert() allows to compare the specific DSDT and TPM2 tables and that's our expectation here I think.

Thanks

Eric

Quoted 18 lines

   Stefan

Quoted 11 lines

Thank you in advance

Best Regards

Eric

Quoted 4 lines

    Stefan

author
#19 lore raw

Re: [RFC 4/6] tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS

Hi Igor,

Quoted 75 lines · Igor Mammedov
Quoted 68 lines · On Fri, 5 Jun 2020 17:47:08 +0200 Auger Eric <[email protected]> wrote:

Hi Stefan,

Quoted 59 lines · Stefan Berger
Quoted 49 lines · Auger Eric

Hi Stefan,

Quoted 35 lines · Stefan Berger
Quoted 34 lines · Auger Eric

Hi Stefan,

Quoted 31 lines · Stefan Berger
Quoted 18 lines · Eric Auger

While writing tests for checking the content of TPM2 and DSDT along with TPM-TIS instantiation I attempted to reuse the framework used for TPM-TIS tests. However While dumping the ACPI tables I get an assert on TPM2_ST_NO_SESSIONS. My assumption is maybe the other tests did not execute long enough to encounter this. So I tentatively propose to remove the assert as it does not seem to break other tests and enable the new ones.

---
    tests/qtest/tpm-emu.c | 1 -

    1 file changed, 1 deletion(-)
tests/qtest/tpm-emu.c +0 -0
index c43ac4aef8..298d0eec74 100644

tests/qtest/tpm-emu.c static void *tpm_emu_tpm_thread(void *data) · line 52

Quoted 4 lines
tests/qtest/tpm-emu.c +0 -1
@@ -49,7 +49,6 @@ static void *tpm_emu_tpm_thread(void *data)
             s->tpm_msg->tag = be16_to_cpu(s->tpm_msg->tag);
             s->tpm_msg->len = be32_to_cpu(s->tpm_msg->len);
             g_assert_cmpint(s->tpm_msg->len, >=, minhlen);
-        g_assert_cmpint(s->tpm_msg->tag, ==, TPM2_ST_NO_SESSIONS);  

You should not have to remove this. The tests are skipped if swtpm does
not support TPM 2 via --tpm2 option. This would be a very old swtpm version, though. So, all tests are run with --tpm2 option and any response received from the TPM would be a TPM 2 response that should have TPM2_ST_NO_SESSIONS as the tag. I'd be curious what other value you
are seeing there.

Quoted 2 lines
Quoted 1 line

If I revert this patch I am getting TPM2_ST_SESSIONS on my end.

Is firmware/BIOS active? There's no TPM2_ST_SESSIONS coming out of QEMU.

So it looks SeaBIOS is in use (bios-256k.bin loaded).

I can see MMIO accesses to the TPM and the following commands are observable:
tpm_emu_tpm_thread code=0x181 tag=0x8001 len=0xa tpm_emu_tpm_thread code=0x144 tag=0x8001 len=0xc tpm_emu_tpm_thread code=0x121 tag=0x8002 len=0x20
This last one causes the assert (TPM2_CC_HierarchyControl)

Quoted 8 lines

I checked in Seabios and effectively tpm20_hierarchycontrol() tags the TPM2_CC_HierarchyControl command with TPM2_ST_SESSIONS

Due to our emulation, maybe tpm_set_failure() gets called, inducing tpm20_hierarchycontrol() call.

That being said, what do you recommend? Remove the assert, improve the emulation, other?

Quoted 6 lines

So this is an ACPI test. What role does the firmware play for success of the test? If the test relies on the firmware showing some sort of expected result, then I would recommend only running this test with an attached swtpm, like we run some other tests. If we don't need the firmware to succeed then I would just get rid of the assert. Probably no other test we have implemented so far was running the firmware...

Quoted 3 lines

FWIU The goal of this test is to compare the acpi tables generated by qemu against reference ones. I dont think we expect from the FW any specific result but I would prefer Igor or Michael to confirm.

Quoted 5 lines

Firmware is needed to fetch tables from QEMU and place them in guest RAM, it will also patch cross table pointers accordingly.

So bios-tables-test checks both QEMU and firmware at the same time, and reference tables are in form guest OS will see them.

Thank you for your input. I guess Stefan's concern was: is that an issue if the FW accesses to the TPM silently fail at some (assumed late) point. I guess the job we expect from the FW, ie. copy the ACPI tables and patch tables, is done prior to that failure, hence the test success. So this functional failure should not be an issue for bios-table-test, right?

Thanks

Eric

Quoted 28 lines
Quoted 26 lines

In that case, removing the assert() allows to compare the specific DSDT and TPM2 tables and that's our expectation here I think.

Thanks

Eric

Quoted 18 lines

   Stefan

Quoted 11 lines

Thank you in advance

Best Regards

Eric

Quoted 4 lines

    Stefan

author
#20 lore raw

RFC 5/6 bios-tables-test: Add Q35/TPM-TIS test

In reply to Eric Auger

Test tables specific to the TPM-TIS instantiation. The TPM2 is added in the framework. Also the DSDT is updated with the TPM. The new function should be be usable for CRB as well, later one.

---
 tests/qtest/bios-tables-test.c | 60 ++++++++++++++++++++++++++++++++++
 tests/qtest/Makefile.include   |  1 +
 2 files changed, 61 insertions(+)
tests/qtest/bios-tables-test.c +60 -0
index c9843829b3..bbba98342c 100644
@@ -57,6 +57,9 @@
 #include "qemu/bitmap.h"
 #include "acpi-utils.h"
 #include "boot-sector.h"
+#include "tpm-emu.h"
+#include "hw/acpi/tpm.h"
+
 
 #define MACHINE_PC "pc"
 #define MACHINE_Q35 "q35"
@@ -874,6 +877,60 @@ static void test_acpi_piix4_tcg_numamem(void)
     free_test_data(&data);
 }
 
+uint64_t tpm_tis_base_addr;
+
+struct tpm_test_data {
+    const char *machine;
+    const char *tpm_if;
+};
+
+static void test_acpi_tcg_tpm(const void *context)
+{
+    struct tpm_test_data *c = (struct tpm_test_data *)context;
+    gchar *tmp_dir_name = g_strdup_printf("qemu-test_acpi_%s_tcg_%s.XXXXXX",
+                                          c->machine, c->tpm_if);
+    char *tmp_path = g_dir_make_tmp(tmp_dir_name, NULL);
+    TestState test;
+    test_data data;
+    GThread *thread;
+    char *args, *variant = g_strdup_printf(".%s", c->tpm_if);
+
+    tpm_tis_base_addr = TPM_TIS_ADDR_BASE;
+
+    module_call_init(MODULE_INIT_QOM);
+
+    test.addr = g_new0(SocketAddress, 1);
+    test.addr->type = SOCKET_ADDRESS_TYPE_UNIX;
+    test.addr->u.q_unix.path = g_build_filename(tmp_path, "sock", NULL);
+    g_mutex_init(&test.data_mutex);
+    g_cond_init(&test.data_cond);
+    test.data_cond_signal = false;
+
+    thread = g_thread_new(NULL, tpm_emu_ctrl_thread, &test);
+    tpm_emu_test_wait_cond(&test);
+
+    memset(&data, 0, sizeof(data));
+    data.machine = c->machine;
+    data.variant = variant;
+
+    args = g_strdup_printf(
+        " -chardev socket,id=chr,path=%s"
+        " -tpmdev emulator,id=dev,chardev=chr"
+        " -device tpm-%s,tpmdev=dev",
+        test.addr->u.q_unix.path, c->tpm_if);
+
+    test_acpi_one(args, &data);
+
+    g_thread_join(thread);
+    g_unlink(test.addr->u.q_unix.path);
+    qapi_free_SocketAddress(test.addr);
+    g_rmdir(tmp_path);
+    g_free(variant);
+    g_free(tmp_path);
+    g_free(tmp_dir_name);
+    free_test_data(&data);
+}
+
 static void test_acpi_tcg_dimm_pxm(const char *machine)
 {
     test_data data;
@@ -1028,6 +1085,7 @@ int main(int argc, char *argv[])
 {
     const char *arch = qtest_get_arch();
     int ret;
+    struct tpm_test_data tpm_q35_tis = {MACHINE_Q35, "tis"};
 
     g_test_init(&argc, &argv, NULL);
 
@@ -1037,6 +1095,8 @@ int main(int argc, char *argv[])
             return ret;
         }
 
+        qtest_add_data_func("acpi/q35/tpm-tis",
+                            &tpm_q35_tis, test_acpi_tcg_tpm);
         qtest_add_func("acpi/piix4", test_acpi_piix4_tcg);
         qtest_add_func("acpi/piix4/bridge", test_acpi_piix4_tcg_bridge);
         qtest_add_func("acpi/q35", test_acpi_q35_tcg);
tests/qtest/Makefile.include +1 -0
index 9e5a51d033..5023fa413d 100644
@@ -262,6 +262,7 @@ tests/qtest/hd-geo-test$(EXESUF): tests/qtest/hd-geo-test.o $(libqos-obj-y)
 tests/qtest/boot-order-test$(EXESUF): tests/qtest/boot-order-test.o $(libqos-obj-y)
 tests/qtest/boot-serial-test$(EXESUF): tests/qtest/boot-serial-test.o $(libqos-obj-y)
 tests/qtest/bios-tables-test$(EXESUF): tests/qtest/bios-tables-test.o \
+        tests/qtest/tpm-emu.o $(test-io-obj-y) \
 	tests/qtest/boot-sector.o tests/qtest/acpi-utils.o $(libqos-obj-y)
 tests/qtest/pxe-test$(EXESUF): tests/qtest/pxe-test.o tests/qtest/boot-sector.o $(libqos-obj-y)
 tests/qtest/microbit-test$(EXESUF): tests/qtest/microbit-test.o
Signature · 1 line
2.20.1
Reviewed-by on v1 5/6
#21 lore raw

Re: [RFC 5/6] bios-tables-test: Add Q35/TPM-TIS test

Quoted 108 lines · Eric Auger

Test tables specific to the TPM-TIS instantiation. The TPM2 is added in the framework. Also the DSDT is updated with the TPM. The new function should be be usable for CRB as well, later one.

---
  tests/qtest/bios-tables-test.c | 60 ++++++++++++++++++++++++++++++++++
  tests/qtest/Makefile.include   |  1 +
  2 files changed, 61 insertions(+)
tests/qtest/bios-tables-test.c +60 -0
index c9843829b3..bbba98342c 100644
@@ -57,6 +57,9 @@
  #include "qemu/bitmap.h"
  #include "acpi-utils.h"
  #include "boot-sector.h"
+#include "tpm-emu.h"
+#include "hw/acpi/tpm.h"
+
 
  #define MACHINE_PC "pc"
  #define MACHINE_Q35 "q35"
@@ -874,6 +877,60 @@ static void test_acpi_piix4_tcg_numamem(void)
      free_test_data(&data);
  }
 
+uint64_t tpm_tis_base_addr;
+
+struct tpm_test_data {
+    const char *machine;
+    const char *tpm_if;
+};
+
+static void test_acpi_tcg_tpm(const void *context)
+{
+    struct tpm_test_data *c = (struct tpm_test_data *)context;
+    gchar *tmp_dir_name = g_strdup_printf("qemu-test_acpi_%s_tcg_%s.XXXXXX",
+                                          c->machine, c->tpm_if);
+    char *tmp_path = g_dir_make_tmp(tmp_dir_name, NULL);
+    TestState test;
+    test_data data;
+    GThread *thread;
+    char *args, *variant = g_strdup_printf(".%s", c->tpm_if);
+
+    tpm_tis_base_addr = TPM_TIS_ADDR_BASE;
+
+    module_call_init(MODULE_INIT_QOM);
+
+    test.addr = g_new0(SocketAddress, 1);
+    test.addr->type = SOCKET_ADDRESS_TYPE_UNIX;
+    test.addr->u.q_unix.path = g_build_filename(tmp_path, "sock", NULL);
+    g_mutex_init(&test.data_mutex);
+    g_cond_init(&test.data_cond);
+    test.data_cond_signal = false;
+
+    thread = g_thread_new(NULL, tpm_emu_ctrl_thread, &test);
+    tpm_emu_test_wait_cond(&test);
+
+    memset(&data, 0, sizeof(data));
+    data.machine = c->machine;
+    data.variant = variant;
+
+    args = g_strdup_printf(
+        " -chardev socket,id=chr,path=%s"
+        " -tpmdev emulator,id=dev,chardev=chr"
+        " -device tpm-%s,tpmdev=dev",
+        test.addr->u.q_unix.path, c->tpm_if);
+
+    test_acpi_one(args, &data);
+
+    g_thread_join(thread);
+    g_unlink(test.addr->u.q_unix.path);
+    qapi_free_SocketAddress(test.addr);
+    g_rmdir(tmp_path);
+    g_free(variant);
+    g_free(tmp_path);
+    g_free(tmp_dir_name);
+    free_test_data(&data);
+}
+
  static void test_acpi_tcg_dimm_pxm(const char *machine)
  {
      test_data data;
@@ -1028,6 +1085,7 @@ int main(int argc, char *argv[])
  {
      const char *arch = qtest_get_arch();
      int ret;
+    struct tpm_test_data tpm_q35_tis = {MACHINE_Q35, "tis"};
 
      g_test_init(&argc, &argv, NULL);
 
@@ -1037,6 +1095,8 @@ int main(int argc, char *argv[])
              return ret;
          }
 
+        qtest_add_data_func("acpi/q35/tpm-tis",
+                            &tpm_q35_tis, test_acpi_tcg_tpm);
          qtest_add_func("acpi/piix4", test_acpi_piix4_tcg);
          qtest_add_func("acpi/piix4/bridge", test_acpi_piix4_tcg_bridge);
          qtest_add_func("acpi/q35", test_acpi_q35_tcg);
tests/qtest/Makefile.include +0 -0
index 9e5a51d033..5023fa413d 100644

tests/qtest/Makefile.include tests/qtest/hd-geo-test$(EXESUF): tests/qtest/hd-geo-test.o $(libqos-obj-y) · line 268

Quoted 7 lines
tests/qtest/Makefile.include +1 -0
@@ -262,6 +262,7 @@ tests/qtest/hd-geo-test$(EXESUF): tests/qtest/hd-geo-test.o $(libqos-obj-y)
  tests/qtest/boot-order-test$(EXESUF): tests/qtest/boot-order-test.o $(libqos-obj-y)
  tests/qtest/boot-serial-test$(EXESUF): tests/qtest/boot-serial-test.o $(libqos-obj-y)
  tests/qtest/bios-tables-test$(EXESUF): tests/qtest/bios-tables-test.o \
+        tests/qtest/tpm-emu.o $(test-io-obj-y) \
  	tests/qtest/boot-sector.o tests/qtest/acpi-utils.o $(libqos-obj-y)
  tests/qtest/pxe-test$(EXESUF): tests/qtest/pxe-test.o tests/qtest/boot-sector.o $(libqos-obj-y)
  tests/qtest/microbit-test$(EXESUF): tests/qtest/microbit-test.o
#22 lore raw

Re: [RFC 5/6] bios-tables-test: Add Q35/TPM-TIS test

In reply to Eric Auger
Quoted 28 lines · On Mon, 1 Jun 2020 12:21:12 +0200 Eric Auger <[email protected]> wrote:

Test tables specific to the TPM-TIS instantiation. The TPM2 is added in the framework. Also the DSDT is updated with the TPM. The new function should be be usable for CRB as well, later one.

---
 tests/qtest/bios-tables-test.c | 60 ++++++++++++++++++++++++++++++++++
 tests/qtest/Makefile.include   |  1 +
 2 files changed, 61 insertions(+)
tests/qtest/bios-tables-test.c +3 -0
index c9843829b3..bbba98342c 100644
@@ -57,6 +57,9 @@
 #include "qemu/bitmap.h"
 #include "acpi-utils.h"
 #include "boot-sector.h"
+#include "tpm-emu.h"
+#include "hw/acpi/tpm.h"
+
 
 #define MACHINE_PC "pc"
 #define MACHINE_Q35 "q35"
@@ -874,6 +877,60 @@ static void test_acpi_piix4_tcg_numamem(void)
     free_test_data(&data);
 }

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 888

Quoted 10 lines
tests/qtest/bios-tables-test.c +9 -0
 
+uint64_t tpm_tis_base_addr;
+
+struct tpm_test_data {
+    const char *machine;
+    const char *tpm_if;
+};
+
+static void test_acpi_tcg_tpm(const void *context)
+{

s/test_acpi_tcg_tpm/test_acpi_q35_tcg_tpm/

I'd try to keep test specific parameter within test function isnstead of pushing it up to main(),
drawback would be some code duplication for intializing test data and calling runner
but it's trivial and worked well so far. See for example test_acpi_piix4_tcg_bridge/test_acpi_q35_tcg_bridge. I might seem a waste but it's consictent with what we were doing with bios tests.

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 896

Quoted 8 lines
tests/qtest/bios-tables-test.c +8 -0
+    struct tpm_test_data *c = (struct tpm_test_data *)context;
+    gchar *tmp_dir_name = g_strdup_printf("qemu-test_acpi_%s_tcg_%s.XXXXXX",
+                                          c->machine, c->tpm_if);
+    char *tmp_path = g_dir_make_tmp(tmp_dir_name, NULL);
+    TestState test;
+    test_data data;
+    GThread *thread;
+    char *args, *variant = g_strdup_printf(".%s", c->tpm_if);

maybe derive tpm_if from '.variant' if it's necessary at all?

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 898

Quoted 2 lines
tests/qtest/bios-tables-test.c +2 -0
+
+    tpm_tis_base_addr = TPM_TIS_ADDR_BASE;

hardcode it here, so in case QEMU regresses, test could notice?

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 900

Quoted 2 lines
tests/qtest/bios-tables-test.c +2 -0
+
+    module_call_init(MODULE_INIT_QOM);

why it's here?

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 910

Quoted 10 lines
tests/qtest/bios-tables-test.c +10 -0
+
+    test.addr = g_new0(SocketAddress, 1);
+    test.addr->type = SOCKET_ADDRESS_TYPE_UNIX;
+    test.addr->u.q_unix.path = g_build_filename(tmp_path, "sock", NULL);
+    g_mutex_init(&test.data_mutex);
+    g_cond_init(&test.data_cond);
+    test.data_cond_signal = false;
+
+    thread = g_thread_new(NULL, tpm_emu_ctrl_thread, &test);
+    tpm_emu_test_wait_cond(&test);

perhaps make a separate helper function from this chunk so it could be reused from other TMP test functions.

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 912

Quoted 2 lines
tests/qtest/bios-tables-test.c +2 -0
+
+    memset(&data, 0, sizeof(data));

I'd init fields with initializer, see test_acpi_virt_tcg_numamem()

Quoted 25 lines
tests/qtest/bios-tables-test.c +21 -0
+    data.machine = c->machine;
+    data.variant = variant;
+
+    args = g_strdup_printf(
+        " -chardev socket,id=chr,path=%s"
+        " -tpmdev emulator,id=dev,chardev=chr"
+        " -device tpm-%s,tpmdev=dev",
+        test.addr->u.q_unix.path, c->tpm_if);
+
+    test_acpi_one(args, &data);
+
+    g_thread_join(thread);
+    g_unlink(test.addr->u.q_unix.path);
+    qapi_free_SocketAddress(test.addr);
+    g_rmdir(tmp_path);
+    g_free(variant);
+    g_free(tmp_path);
+    g_free(tmp_dir_name);
+    free_test_data(&data);
+}
+
 static void test_acpi_tcg_dimm_pxm(const char *machine)
 {
     test_data data;

tests/qtest/bios-tables-test.c int main(int argc, char *argv[]) · line 1088

Quoted 4 lines
tests/qtest/bios-tables-test.c +1 -0
@@ -1028,6 +1085,7 @@ int main(int argc, char *argv[])
 {
     const char *arch = qtest_get_arch();
     int ret;
+    struct tpm_test_data tpm_q35_tis = {MACHINE_Q35, "tis"};

I'd hide this within test_acpi_tcg_tpm() as it's done in other test functions

Quoted 24 lines
tests/qtest/bios-tables-test.c +2 -0
     g_test_init(&argc, &argv, NULL);
 
@@ -1037,6 +1095,8 @@ int main(int argc, char *argv[])
             return ret;
         }
 
+        qtest_add_data_func("acpi/q35/tpm-tis",
+                            &tpm_q35_tis, test_acpi_tcg_tpm);
         qtest_add_func("acpi/piix4", test_acpi_piix4_tcg);
         qtest_add_func("acpi/piix4/bridge", test_acpi_piix4_tcg_bridge);
         qtest_add_func("acpi/q35", test_acpi_q35_tcg);
tests/qtest/Makefile.include +1 -0
index 9e5a51d033..5023fa413d 100644
@@ -262,6 +262,7 @@ tests/qtest/hd-geo-test$(EXESUF): tests/qtest/hd-geo-test.o $(libqos-obj-y)
 tests/qtest/boot-order-test$(EXESUF): tests/qtest/boot-order-test.o $(libqos-obj-y)
 tests/qtest/boot-serial-test$(EXESUF): tests/qtest/boot-serial-test.o $(libqos-obj-y)
 tests/qtest/bios-tables-test$(EXESUF): tests/qtest/bios-tables-test.o \
+        tests/qtest/tpm-emu.o $(test-io-obj-y) \
 	tests/qtest/boot-sector.o tests/qtest/acpi-utils.o $(libqos-obj-y)
 tests/qtest/pxe-test$(EXESUF): tests/qtest/pxe-test.o tests/qtest/boot-sector.o $(libqos-obj-y)
 tests/qtest/microbit-test$(EXESUF): tests/qtest/microbit-test.o
author
#23 lore raw

Re: [RFC 5/6] bios-tables-test: Add Q35/TPM-TIS test

Hi Igor,

Quoted 44 lines · Igor Mammedov
Quoted 28 lines · On Mon, 1 Jun 2020 12:21:12 +0200 Eric Auger <[email protected]> wrote:

Test tables specific to the TPM-TIS instantiation. The TPM2 is added in the framework. Also the DSDT is updated with the TPM. The new function should be be usable for CRB as well, later one.

---
 tests/qtest/bios-tables-test.c | 60 ++++++++++++++++++++++++++++++++++
 tests/qtest/Makefile.include   |  1 +
 2 files changed, 61 insertions(+)
tests/qtest/bios-tables-test.c +3 -0
index c9843829b3..bbba98342c 100644
@@ -57,6 +57,9 @@
 #include "qemu/bitmap.h"
 #include "acpi-utils.h"
 #include "boot-sector.h"
+#include "tpm-emu.h"
+#include "hw/acpi/tpm.h"
+
 
 #define MACHINE_PC "pc"
 #define MACHINE_Q35 "q35"
@@ -874,6 +877,60 @@ static void test_acpi_piix4_tcg_numamem(void)
     free_test_data(&data);
 }

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 888

Quoted 10 lines
tests/qtest/bios-tables-test.c +9 -0
 
+uint64_t tpm_tis_base_addr;
+
+struct tpm_test_data {
+    const char *machine;
+    const char *tpm_if;
+};
+
+static void test_acpi_tcg_tpm(const void *context)
+{

s/test_acpi_tcg_tpm/test_acpi_q35_tcg_tpm/

Quoted 4 lines

I'd try to keep test specific parameter within test function isnstead of pushing it up to main(),
drawback would be some code duplication for intializing test data and calling runner
but it's trivial and worked well so far. See for example test_acpi_piix4_tcg_bridge/test_acpi_q35_tcg_bridge. I might seem a waste but it's consictent with what we were doing with bios tests.

OK

Quoted 12 lines

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 896

Quoted 10 lines
tests/qtest/bios-tables-test.c +8 -0
+    struct tpm_test_data *c = (struct tpm_test_data *)context;
+    gchar *tmp_dir_name = g_strdup_printf("qemu-test_acpi_%s_tcg_%s.XXXXXX",
+                                          c->machine, c->tpm_if);
+    char *tmp_path = g_dir_make_tmp(tmp_dir_name, NULL);
+    TestState test;
+    test_data data;
+    GThread *thread;
+    char *args, *variant = g_strdup_printf(".%s", c->tpm_if);

maybe derive tpm_if from '.variant' if it's necessary at all?

Quoted 7 lines

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 898

Quoted 2 lines
tests/qtest/bios-tables-test.c +2 -0
+
+    tpm_tis_base_addr = TPM_TIS_ADDR_BASE;

hardcode it here, so in case QEMU regresses, test could notice?

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 900

Quoted 2 lines
tests/qtest/bios-tables-test.c +2 -0
+
+    module_call_init(MODULE_INIT_QOM);

why it's here

Without it, I get
/x86_64/acpi/q35/tpm-tis: ** ERROR:qom/object.c:677:object_new_with_type: assertion failed: (type != NULL)

Thanks

Eric

Quoted 75 lines

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 910

Quoted 11 lines
tests/qtest/bios-tables-test.c +10 -0
+
+    test.addr = g_new0(SocketAddress, 1);
+    test.addr->type = SOCKET_ADDRESS_TYPE_UNIX;
+    test.addr->u.q_unix.path = g_build_filename(tmp_path, "sock", NULL);
+    g_mutex_init(&test.data_mutex);
+    g_cond_init(&test.data_cond);
+    test.data_cond_signal = false;
+
+    thread = g_thread_new(NULL, tpm_emu_ctrl_thread, &test);
+    tpm_emu_test_wait_cond(&test);

perhaps make a separate helper function from this chunk so it could be reused from other TMP test functions.

tests/qtest/bios-tables-test.c static void test_acpi_piix4_tcg_numamem(void) · line 912

Quoted 2 lines
tests/qtest/bios-tables-test.c +2 -0
+
+    memset(&data, 0, sizeof(data));

I'd init fields with initializer, see test_acpi_virt_tcg_numamem()

Quoted 25 lines
tests/qtest/bios-tables-test.c +21 -0
+    data.machine = c->machine;
+    data.variant = variant;
+
+    args = g_strdup_printf(
+        " -chardev socket,id=chr,path=%s"
+        " -tpmdev emulator,id=dev,chardev=chr"
+        " -device tpm-%s,tpmdev=dev",
+        test.addr->u.q_unix.path, c->tpm_if);
+
+    test_acpi_one(args, &data);
+
+    g_thread_join(thread);
+    g_unlink(test.addr->u.q_unix.path);
+    qapi_free_SocketAddress(test.addr);
+    g_rmdir(tmp_path);
+    g_free(variant);
+    g_free(tmp_path);
+    g_free(tmp_dir_name);
+    free_test_data(&data);
+}
+
 static void test_acpi_tcg_dimm_pxm(const char *machine)
 {
     test_data data;

tests/qtest/bios-tables-test.c int main(int argc, char *argv[]) · line 1088

Quoted 4 lines
tests/qtest/bios-tables-test.c +1 -0
@@ -1028,6 +1085,7 @@ int main(int argc, char *argv[])
 {
     const char *arch = qtest_get_arch();
     int ret;
+    struct tpm_test_data tpm_q35_tis = {MACHINE_Q35, "tis"};

I'd hide this within test_acpi_tcg_tpm() as it's done in other test functions

Quoted 24 lines
tests/qtest/bios-tables-test.c +2 -0
     g_test_init(&argc, &argv, NULL);
 
@@ -1037,6 +1095,8 @@ int main(int argc, char *argv[])
             return ret;
         }
 
+        qtest_add_data_func("acpi/q35/tpm-tis",
+                            &tpm_q35_tis, test_acpi_tcg_tpm);
         qtest_add_func("acpi/piix4", test_acpi_piix4_tcg);
         qtest_add_func("acpi/piix4/bridge", test_acpi_piix4_tcg_bridge);
         qtest_add_func("acpi/q35", test_acpi_q35_tcg);
tests/qtest/Makefile.include +1 -0
index 9e5a51d033..5023fa413d 100644
@@ -262,6 +262,7 @@ tests/qtest/hd-geo-test$(EXESUF): tests/qtest/hd-geo-test.o $(libqos-obj-y)
 tests/qtest/boot-order-test$(EXESUF): tests/qtest/boot-order-test.o $(libqos-obj-y)
 tests/qtest/boot-serial-test$(EXESUF): tests/qtest/boot-serial-test.o $(libqos-obj-y)
 tests/qtest/bios-tables-test$(EXESUF): tests/qtest/bios-tables-test.o \
+        tests/qtest/tpm-emu.o $(test-io-obj-y) \
 	tests/qtest/boot-sector.o tests/qtest/acpi-utils.o $(libqos-obj-y)
 tests/qtest/pxe-test$(EXESUF): tests/qtest/pxe-test.o tests/qtest/boot-sector.o $(libqos-obj-y)
 tests/qtest/microbit-test$(EXESUF): tests/qtest/microbit-test.o
author
#24 lore raw

RFC 6/6 bios-tables-test: Generate reference tables for Q35/TPM-TIS

In reply to Eric Auger

TPM2, DSDT tables were generated using
tests/data/acpi/rebuild-expected-aml.sh

---
 tests/qtest/bios-tables-test-allowed-diff.h |   2 --
 tests/data/acpi/q35/DSDT.tis                | Bin 0 -> 8468 bytes
 tests/data/acpi/q35/TPM2.tis                | Bin 0 -> 76 bytes
 3 files changed, 2 deletions(-)
tests/qtest/bios-tables-test-allowed-diff.h +0 -2
index a2a45d1d31..dfb8523c8b 100644
@@ -1,3 +1 @@
 /* List of comma-separated changed AML files to ignore */
-"tests/data/acpi/q35/DSDT.tis",
-"tests/data/acpi/q35/TPM2.tis",
tests/data/acpi/q35/DSDT.tis binary +0 -0
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..34a3c915d27cab5d09190a0441e9511db0c147f2 100644
GIT binary patch
tests/data/acpi/q35/TPM2.tis binary +0 -0
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..7878a6e79a6a406d99ca1f5e9a528eb392b8d793 100644
GIT binary patch
Signature · 1 line
2.20.1
Reviewed-by on v1 6/6
#25 lore raw

Re: [RFC 6/6] bios-tables-test: Generate reference tables for Q35/TPM-TIS

Quoted 4 lines · Eric Auger

TPM2, DSDT tables were generated using
tests/data/acpi/rebuild-expected-aml.sh

Quoted 93 lines
---
  tests/qtest/bios-tables-test-allowed-diff.h |   2 --
  tests/data/acpi/q35/DSDT.tis                | Bin 0 -> 8468 bytes
  tests/data/acpi/q35/TPM2.tis                | Bin 0 -> 76 bytes
  3 files changed, 2 deletions(-)
tests/qtest/bios-tables-test-allowed-diff.h +0 -2
index a2a45d1d31..dfb8523c8b 100644
@@ -1,3 +1 @@
  /* List of comma-separated changed AML files to ignore */
-"tests/data/acpi/q35/DSDT.tis",
-"tests/data/acpi/q35/TPM2.tis",
tests/data/acpi/q35/DSDT.tis binary +0 -0
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..34a3c915d27cab5d09190a0441e9511db0c147f2 100644
GIT binary patch
tests/data/acpi/q35/TPM2.tis binary +0 -0
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..7878a6e79a6a406d99ca1f5e9a528eb392b8d793 100644
GIT binary patch
author
#26 lore raw
In reply to Eric Auger

Hi,

Quoted 10 lines · Eric Auger

This series adds the test of the TPM2 and DSDT tables when the TPM-TIS device gets instantiated.

The series also allows to validate changes in the TPM2 table generation done in "[PATCH v3 0/4] vTPM/aarch64 ACPI support", sent separately.

This depends on Stefan's "acpi: tpm: Do not build TCPA table for TPM 2".

Quoted 8 lines

This is in RFC state because I am not very familiar with the table generation process docommented in bios-tables-test.c.

I don't know how to get rid of the checkpatch errors:
"ERROR: Does not appear to be a unified-diff format patch"

Surprisingly, checkpatch argues when void tables are added in the same time as bios-tables-test-allowed-diff.h changes.

Both issues reported above were resolved by Michael's [PATCH] checkpatch: reversed logic with acpi test checks

Thank you for fixing that

Eric

Quoted 31 lines

Most importantly, to be able to reuse tpm-emu code, I was forced to remove the TPM2_ST_NO_SESSIONS assert. I would be grateful if I could get some advises on the best way to address the issue.

Best Regards

Eric

This series can be found at:
https://github.com/eauger/qemu/tree/v5.0-tpm-acpi-tests-v1

Eric Auger (6):
test/tpm-emu: include sockets and channel headers in tpm-emu header tests/acpi: Add void tables for Q35/TPM-TIS bios-tables-test tests/acpi: Ignore TPM2.tis and DSDT.tis
tests: tpm-emu: Remove assert on TPM2_ST_NO_SESSIONS
bios-tables-test: Add Q35/TPM-TIS test
bios-tables-test: Generate reference tables for Q35/TPM-TIS

 tests/qtest/tpm-emu.h          |   3 ++
 tests/qtest/bios-tables-test.c |  60 +++++++++++++++++++++++++++++++++
 tests/qtest/tpm-emu.c          |   1 -
 tests/data/acpi/q35/DSDT.tis   | Bin 0 -> 8468 bytes
 tests/data/acpi/q35/TPM2.tis   | Bin 0 -> 76 bytes
 tests/qtest/Makefile.include   |   1 +
 6 files changed, 64 insertions(+), 1 deletion(-)
 create mode 100644 tests/data/acpi/q35/DSDT.tis
 create mode 100644 tests/data/acpi/q35/TPM2.tis