Re: [PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Jarkko Sakkinen]

On Tue, 2018-03-06 at 15:19 +, Winkler, Tomas wrote:
> > On Tue, 2018-03-06 at 11:25 +0200, Tomas Winkler wrote:
> > > TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve
> > > generation of crypto keys which can be a computationally intensive task.
> > > The timeout is set to 3min.
> > > Rather than increasing default timeout a new constant is added, to not
> > > stall for too long on regular commands failures.
> > > 
> > > Signed-off-by: Tomas Winkler 
> > 
> > Why are you radically chaging the default timeout? The commit message
> > does not tell anything about that change.
> > 
> 
> Let me, recheck but it should be same value just converted to msecs.
>  
> > Why couldn't we just have two timeouts: one default and one long that
> > would be at least as long as the longest timeout defined in the spec?
> 
> I've tried to explain it in the commit message but apparently has failed. 
> 
> We have a default or undefined which should be the same as it was unless I did
> some silly mistake in conversion to msecs (will check), ass all others are in
> msecs.
> It was 2 min = 2 * 60 * HZ (in jiffies) which would be msecs_to_jiffies(2 * 60
> * 1000 = 12) 
> TPM2_DURATION_DEFAULT   = 12

Aah, of course :-) The problem was that I had somehow a blid spot with
seeing the msec_to_jiffies() conversion.

TPM_NUM_DURATIONS would a better name than TPM_DURATION_MAX because
TPM_DURATION_MAX is easy to confuse with TPM_DURATION_* constants.

/Jarkko

Re: [PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Jarkko Sakkinen]

On Tue, 2018-03-06 at 15:19 +, Winkler, Tomas wrote:
> > On Tue, 2018-03-06 at 11:25 +0200, Tomas Winkler wrote:
> > > TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve
> > > generation of crypto keys which can be a computationally intensive task.
> > > The timeout is set to 3min.
> > > Rather than increasing default timeout a new constant is added, to not
> > > stall for too long on regular commands failures.
> > > 
> > > Signed-off-by: Tomas Winkler 
> > 
> > Why are you radically chaging the default timeout? The commit message
> > does not tell anything about that change.
> > 
> 
> Let me, recheck but it should be same value just converted to msecs.
>  
> > Why couldn't we just have two timeouts: one default and one long that
> > would be at least as long as the longest timeout defined in the spec?
> 
> I've tried to explain it in the commit message but apparently has failed. 
> 
> We have a default or undefined which should be the same as it was unless I did
> some silly mistake in conversion to msecs (will check), ass all others are in
> msecs.
> It was 2 min = 2 * 60 * HZ (in jiffies) which would be msecs_to_jiffies(2 * 60
> * 1000 = 12) 
> TPM2_DURATION_DEFAULT   = 12

Aah, of course :-) The problem was that I had somehow a blid spot with
seeing the msec_to_jiffies() conversion.

TPM_NUM_DURATIONS would a better name than TPM_DURATION_MAX because
TPM_DURATION_MAX is easy to confuse with TPM_DURATION_* constants.

/Jarkko

RE: [PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Winkler, Tomas]

> On Tue, 2018-03-06 at 11:25 +0200, Tomas Winkler wrote:
> > TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve
> > generation of crypto keys which can be a computationally intensive task.
> > The timeout is set to 3min.
> > Rather than increasing default timeout a new constant is added, to not
> > stall for too long on regular commands failures.
> >
> > Signed-off-by: Tomas Winkler 
> 
> Why are you radically chaging the default timeout? The commit message
> does not tell anything about that change.
>
Let me, recheck but it should be same value just converted to msecs.
 
> Why couldn't we just have two timeouts: one default and one long that
> would be at least as long as the longest timeout defined in the spec?

I've tried to explain it in the commit message but apparently has failed. 

We have a default or undefined which should be the same as it was unless I did 
some silly mistake in conversion to msecs (will check), ass all others are in 
msecs.
It was 2 min = 2 * 60 * HZ (in jiffies) which would be msecs_to_jiffies(2 * 60 
* 1000 = 12) 
TPM2_DURATION_DEFAULT   = 12

Then we have a new one which is 3 min just for key generation (Prime number 
generation is heavy liffting) 

> Then we would not have to ever update LAST_CC constant, we could actually
> remove it completely.

I believe I've removed it next patch, I didn’t want do too much in this one.

Thanks
Tomas

RE: [PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Winkler, Tomas]

> On Tue, 2018-03-06 at 11:25 +0200, Tomas Winkler wrote:
> > TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve
> > generation of crypto keys which can be a computationally intensive task.
> > The timeout is set to 3min.
> > Rather than increasing default timeout a new constant is added, to not
> > stall for too long on regular commands failures.
> >
> > Signed-off-by: Tomas Winkler 
> 
> Why are you radically chaging the default timeout? The commit message
> does not tell anything about that change.
>
Let me, recheck but it should be same value just converted to msecs.
 
> Why couldn't we just have two timeouts: one default and one long that
> would be at least as long as the longest timeout defined in the spec?

I've tried to explain it in the commit message but apparently has failed. 

We have a default or undefined which should be the same as it was unless I did 
some silly mistake in conversion to msecs (will check), ass all others are in 
msecs.
It was 2 min = 2 * 60 * HZ (in jiffies) which would be msecs_to_jiffies(2 * 60 
* 1000 = 12) 
TPM2_DURATION_DEFAULT   = 12

Then we have a new one which is 3 min just for key generation (Prime number 
generation is heavy liffting) 

> Then we would not have to ever update LAST_CC constant, we could actually
> remove it completely.

I believe I've removed it next patch, I didn’t want do too much in this one.

Thanks
Tomas

Re: [PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Jarkko Sakkinen]

On Tue, 2018-03-06 at 11:25 +0200, Tomas Winkler wrote:
> TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve generation
> of crypto keys which can be a computationally intensive task.
> The timeout is set to 3min.
> Rather than increasing default timeout a new constant is
> added, to not stall for too long on regular commands failures.
> 
> Signed-off-by: Tomas Winkler 

Why are you radically chaging the default timeout? The commit
message does not tell anything about that change.

Why couldn't we just have two timeouts: one default and one long that
would be at least as long as the longest timeout defined in the spec?
Then we would not have to ever update LAST_CC constant, we could
actually remove it completely.

/Jarkko

Re: [PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Jarkko Sakkinen]

On Tue, 2018-03-06 at 11:25 +0200, Tomas Winkler wrote:
> TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve generation
> of crypto keys which can be a computationally intensive task.
> The timeout is set to 3min.
> Rather than increasing default timeout a new constant is
> added, to not stall for too long on regular commands failures.
> 
> Signed-off-by: Tomas Winkler 

Why are you radically chaging the default timeout? The commit
message does not tell anything about that change.

Why couldn't we just have two timeouts: one default and one long that
would be at least as long as the longest timeout defined in the spec?
Then we would not have to ever update LAST_CC constant, we could
actually remove it completely.

/Jarkko

[PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Tomas Winkler]

TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve generation
of crypto keys which can be a computationally intensive task.
The timeout is set to 3min.
Rather than increasing default timeout a new constant is
added, to not stall for too long on regular commands failures.

Signed-off-by: Tomas Winkler 
---
V2: add more explanation to the commit message.

 drivers/char/tpm/tpm-interface.c |  3 +++
 drivers/char/tpm/tpm.h   | 28 ++--
 drivers/char/tpm/tpm2-cmd.c  |  8 +---
 3 files changed, 26 insertions(+), 13 deletions(-)

diff --git a/drivers/char/tpm/tpm-interface.c b/drivers/char/tpm/tpm-interface.c
index 402e54252b22..40d1770f6b38 100644
--- a/drivers/char/tpm/tpm-interface.c
+++ b/drivers/char/tpm/tpm-interface.c
@@ -421,6 +421,8 @@ int tpm_get_timeouts(struct tpm_chip *chip)
msecs_to_jiffies(TPM2_DURATION_MEDIUM);
chip->duration[TPM_LONG] =
msecs_to_jiffies(TPM2_DURATION_LONG);
+   chip->duration[TPM_LONG_LONG] =
+   msecs_to_jiffies(TPM2_DURATION_LONG_LONG);
 
chip->flags |= TPM_CHIP_FLAG_HAVE_TIMEOUTS;
return 0;
@@ -509,6 +511,7 @@ int tpm_get_timeouts(struct tpm_chip *chip)
usecs_to_jiffies(be32_to_cpu(cap.duration.tpm_medium));
chip->duration[TPM_LONG] =
usecs_to_jiffies(be32_to_cpu(cap.duration.tpm_long));
+   chip->duration[TPM_LONG_LONG] = 0; /* not used under 1.2 */
 
/* The Broadcom BCM0102 chipset in a Dell Latitude D820 gets the above
 * value wrong and apparently reports msecs rather than usecs. So we
diff --git a/drivers/char/tpm/tpm.h b/drivers/char/tpm/tpm.h
index 38197a30ad7b..30610d97d30c 100644
--- a/drivers/char/tpm/tpm.h
+++ b/drivers/char/tpm/tpm.h
@@ -67,7 +67,9 @@ enum tpm_duration {
TPM_SHORT = 0,
TPM_MEDIUM = 1,
TPM_LONG = 2,
+   TPM_LONG_LONG = 3,
TPM_UNDEFINED,
+   TPM_DURATION_MAX = TPM_UNDEFINED,
 };
 
 #define TPM_WARN_RETRY  0x800
@@ -79,15 +81,20 @@ enum tpm_duration {
 #define TPM_HEADER_SIZE10
 
 enum tpm2_const {
-   TPM2_PLATFORM_PCR   = 24,
-   TPM2_PCR_SELECT_MIN = ((TPM2_PLATFORM_PCR + 7) / 8),
-   TPM2_TIMEOUT_A  = 750,
-   TPM2_TIMEOUT_B  = 2000,
-   TPM2_TIMEOUT_C  = 200,
-   TPM2_TIMEOUT_D  = 30,
-   TPM2_DURATION_SHORT = 20,
-   TPM2_DURATION_MEDIUM= 750,
-   TPM2_DURATION_LONG  = 2000,
+   TPM2_PLATFORM_PCR   = 24,
+   TPM2_PCR_SELECT_MIN = ((TPM2_PLATFORM_PCR + 7) / 8),
+};
+
+enum tpm2_timeouts {
+   TPM2_TIMEOUT_A  =750,
+   TPM2_TIMEOUT_B  =   2000,
+   TPM2_TIMEOUT_C  =200,
+   TPM2_TIMEOUT_D  = 30,
+   TPM2_DURATION_SHORT = 20,
+   TPM2_DURATION_MEDIUM=750,
+   TPM2_DURATION_LONG  =   2000,
+   TPM2_DURATION_LONG_LONG = 30,
+   TPM2_DURATION_DEFAULT   = 12,
 };
 
 enum tpm2_structures {
@@ -123,6 +130,7 @@ enum tpm2_algorithms {
 
 enum tpm2_command_codes {
TPM2_CC_FIRST   = 0x011F,
+   TPM2_CC_CREATE_PRIMARY  = 0x0131,
TPM2_CC_SELF_TEST   = 0x0143,
TPM2_CC_STARTUP = 0x0144,
TPM2_CC_SHUTDOWN= 0x0145,
@@ -227,7 +235,7 @@ struct tpm_chip {
unsigned long timeout_c; /* jiffies */
unsigned long timeout_d; /* jiffies */
bool timeout_adjusted;
-   unsigned long duration[3]; /* jiffies */
+   unsigned long duration[TPM_DURATION_MAX]; /* jiffies */
bool duration_adjusted;
 
struct dentry *bios_dir[TPM_NUM_EVENT_LOG_FILES];
diff --git a/drivers/char/tpm/tpm2-cmd.c b/drivers/char/tpm/tpm2-cmd.c
index a700f8f9ead7..c1ddbbba406e 100644
--- a/drivers/char/tpm/tpm2-cmd.c
+++ b/drivers/char/tpm/tpm2-cmd.c
@@ -90,6 +90,8 @@ static struct tpm2_hash tpm2_hash_map[] = {
  * of time the chip could take to return the result. The values
  * of the SHORT, MEDIUM, and LONG durations are taken from the
  * PC Client Profile (PTP) specification.
+ * LONG_LONG is for commands that generates keys which empirically
+ * takes longer time on some systems.
  */
 static const u8 tpm2_ordinal_duration[TPM2_CC_LAST - TPM2_CC_FIRST + 1] = {
TPM_UNDEFINED,  /* 11F */
@@ -110,7 +112,7 @@ static const u8 tpm2_ordinal_duration[TPM2_CC_LAST - 
TPM2_CC_FIRST + 1] = {
TPM_UNDEFINED,  /* 12e */
TPM_UNDEFINED,  /* 12f */
TPM_UNDEFINED,  /* 130 */
-   TPM_UNDEFINED,  /* 131 */
+   TPM_LONG_LONG,  /* 131 */
TPM_UNDEFINED,  /* 132 */
TPM_UNDEFINED,  /* 133 */
TPM_UNDEFINED,  /* 134 */
@@ -144,7 +146,7 @@ static const u8 tpm2_ordinal_duration[TPM2_CC_LAST - 
TPM2_CC_FIRST + 1] = {
TPM_UNDEFINED,  /* 150 */

[PATCH 3/5 V2] tpm2: add longer timeouts for creation commands.

[Tomas Winkler]

TPM2_CC_Create(0x153) and TPM2_CC_CreatePrimary (0x131) involve generation
of crypto keys which can be a computationally intensive task.
The timeout is set to 3min.
Rather than increasing default timeout a new constant is
added, to not stall for too long on regular commands failures.

Signed-off-by: Tomas Winkler 
---
V2: add more explanation to the commit message.

 drivers/char/tpm/tpm-interface.c |  3 +++
 drivers/char/tpm/tpm.h   | 28 ++--
 drivers/char/tpm/tpm2-cmd.c  |  8 +---
 3 files changed, 26 insertions(+), 13 deletions(-)

diff --git a/drivers/char/tpm/tpm-interface.c b/drivers/char/tpm/tpm-interface.c
index 402e54252b22..40d1770f6b38 100644
--- a/drivers/char/tpm/tpm-interface.c
+++ b/drivers/char/tpm/tpm-interface.c
@@ -421,6 +421,8 @@ int tpm_get_timeouts(struct tpm_chip *chip)
msecs_to_jiffies(TPM2_DURATION_MEDIUM);
chip->duration[TPM_LONG] =
msecs_to_jiffies(TPM2_DURATION_LONG);
+   chip->duration[TPM_LONG_LONG] =
+   msecs_to_jiffies(TPM2_DURATION_LONG_LONG);
 
chip->flags |= TPM_CHIP_FLAG_HAVE_TIMEOUTS;
return 0;
@@ -509,6 +511,7 @@ int tpm_get_timeouts(struct tpm_chip *chip)
usecs_to_jiffies(be32_to_cpu(cap.duration.tpm_medium));
chip->duration[TPM_LONG] =
usecs_to_jiffies(be32_to_cpu(cap.duration.tpm_long));
+   chip->duration[TPM_LONG_LONG] = 0; /* not used under 1.2 */
 
/* The Broadcom BCM0102 chipset in a Dell Latitude D820 gets the above
 * value wrong and apparently reports msecs rather than usecs. So we
diff --git a/drivers/char/tpm/tpm.h b/drivers/char/tpm/tpm.h
index 38197a30ad7b..30610d97d30c 100644
--- a/drivers/char/tpm/tpm.h
+++ b/drivers/char/tpm/tpm.h
@@ -67,7 +67,9 @@ enum tpm_duration {
TPM_SHORT = 0,
TPM_MEDIUM = 1,
TPM_LONG = 2,
+   TPM_LONG_LONG = 3,
TPM_UNDEFINED,
+   TPM_DURATION_MAX = TPM_UNDEFINED,
 };
 
 #define TPM_WARN_RETRY  0x800
@@ -79,15 +81,20 @@ enum tpm_duration {
 #define TPM_HEADER_SIZE10
 
 enum tpm2_const {
-   TPM2_PLATFORM_PCR   = 24,
-   TPM2_PCR_SELECT_MIN = ((TPM2_PLATFORM_PCR + 7) / 8),
-   TPM2_TIMEOUT_A  = 750,
-   TPM2_TIMEOUT_B  = 2000,
-   TPM2_TIMEOUT_C  = 200,
-   TPM2_TIMEOUT_D  = 30,
-   TPM2_DURATION_SHORT = 20,
-   TPM2_DURATION_MEDIUM= 750,
-   TPM2_DURATION_LONG  = 2000,
+   TPM2_PLATFORM_PCR   = 24,
+   TPM2_PCR_SELECT_MIN = ((TPM2_PLATFORM_PCR + 7) / 8),
+};
+
+enum tpm2_timeouts {
+   TPM2_TIMEOUT_A  =750,
+   TPM2_TIMEOUT_B  =   2000,
+   TPM2_TIMEOUT_C  =200,
+   TPM2_TIMEOUT_D  = 30,
+   TPM2_DURATION_SHORT = 20,
+   TPM2_DURATION_MEDIUM=750,
+   TPM2_DURATION_LONG  =   2000,
+   TPM2_DURATION_LONG_LONG = 30,
+   TPM2_DURATION_DEFAULT   = 12,
 };
 
 enum tpm2_structures {
@@ -123,6 +130,7 @@ enum tpm2_algorithms {
 
 enum tpm2_command_codes {
TPM2_CC_FIRST   = 0x011F,
+   TPM2_CC_CREATE_PRIMARY  = 0x0131,
TPM2_CC_SELF_TEST   = 0x0143,
TPM2_CC_STARTUP = 0x0144,
TPM2_CC_SHUTDOWN= 0x0145,
@@ -227,7 +235,7 @@ struct tpm_chip {
unsigned long timeout_c; /* jiffies */
unsigned long timeout_d; /* jiffies */
bool timeout_adjusted;
-   unsigned long duration[3]; /* jiffies */
+   unsigned long duration[TPM_DURATION_MAX]; /* jiffies */
bool duration_adjusted;
 
struct dentry *bios_dir[TPM_NUM_EVENT_LOG_FILES];
diff --git a/drivers/char/tpm/tpm2-cmd.c b/drivers/char/tpm/tpm2-cmd.c
index a700f8f9ead7..c1ddbbba406e 100644
--- a/drivers/char/tpm/tpm2-cmd.c
+++ b/drivers/char/tpm/tpm2-cmd.c
@@ -90,6 +90,8 @@ static struct tpm2_hash tpm2_hash_map[] = {
  * of time the chip could take to return the result. The values
  * of the SHORT, MEDIUM, and LONG durations are taken from the
  * PC Client Profile (PTP) specification.
+ * LONG_LONG is for commands that generates keys which empirically
+ * takes longer time on some systems.
  */
 static const u8 tpm2_ordinal_duration[TPM2_CC_LAST - TPM2_CC_FIRST + 1] = {
TPM_UNDEFINED,  /* 11F */
@@ -110,7 +112,7 @@ static const u8 tpm2_ordinal_duration[TPM2_CC_LAST - 
TPM2_CC_FIRST + 1] = {
TPM_UNDEFINED,  /* 12e */
TPM_UNDEFINED,  /* 12f */
TPM_UNDEFINED,  /* 130 */
-   TPM_UNDEFINED,  /* 131 */
+   TPM_LONG_LONG,  /* 131 */
TPM_UNDEFINED,  /* 132 */
TPM_UNDEFINED,  /* 133 */
TPM_UNDEFINED,  /* 134 */
@@ -144,7 +146,7 @@ static const u8 tpm2_ordinal_duration[TPM2_CC_LAST - 
TPM2_CC_FIRST + 1] = {
TPM_UNDEFINED,  /* 150 */
TPM_UNDEFINED,  /*