TEE: minor cleanup
Subsystems (2)
v4 cover TEE: minor cleanup
- Address some spelling errors and typos
- Support CMD_OPTEE_RPMB for SANDBOX configurations and add python tests
- Remove common.h inclusion for drivers/tee
- Add calls for closing tee session after every read/write operation.
CI build:
[1] https://dev.azure.com/igoropaniuk/u-boot/_build/results?buildId=31&view=results
Changes in v4:
- Rebased on the latest master and excluded "tee: sandbox: fix spelling errors", as it was merged already by Heinrich Schuchardt
Changes in v3:
- Added calls for closing tee session after every read/write operation
Changes in v2:
- Fixed chimp_optee.c:37:9: error: implicit declaration of function 'memset'
- Applied R-b and T-b tags
Igor Opaniuk (5):
tee: optee: fix description in Kconfig
cmd: optee_rpmb: close tee session
cmd: optee_rpmb: build cmd for sandbox
test: py: add optee_rpmb tests
tee: remove common.h inclusion
cmd/Kconfig | 4 +++- cmd/optee_rpmb.c | 23 +++++++++++++++++------ drivers/tee/broadcom/chimp_optee.c | 3 ++- drivers/tee/optee/Kconfig | 2 +- drivers/tee/optee/core.c | 1 - drivers/tee/optee/i2c.c | 1 - drivers/tee/optee/rpmb.c | 1 - drivers/tee/optee/supplicant.c | 2 +- drivers/tee/sandbox.c | 2 +- drivers/tee/tee-uclass.c | 1 - test/py/tests/test_optee_rpmb.py | 20 ++++++++++++++++++++ 11 files changed, 45 insertions(+), 15 deletions(-) create mode 100644 test/py/tests/test_optee_rpmb.py
Signature · 1 line
2.34.1
PATCH v4 1/5 tee: optee: fix description in Kconfig
Fix OPTEE_TA_AVB symbol description in Kconfig:
s/"write"rb"/"write_rb"/g
---
(no changes since v2)
Changes in v2:
- Applied R-b tags
drivers/tee/optee/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/tee/optee/Kconfig +1 -1
@@ -19,7 +19,7 @@ config OPTEE_TA_AVB default y help Enables support for the AVB Trusted Application (TA) in OP-TEE. - The TA can support the "avb" subcommands "read_rb", "write"rb" + The TA can support the "avb" subcommands "read_rb", "write_rb" and "is_unlocked". config OPTEE_TA_RPC_TEST
Signature · 1 line
2.34.1
PATCH v4 2/5 cmd: optee_rpmb: close tee session
In reply to Igor OpaniukAdd calls for closing tee session after every read/write operation.
---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)
cmd/optee_rpmb.c +17 -6
@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name, rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + } rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);
@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }
@@ -139,17 +144,20 @@ static int write_persistent_value(const char *name,struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size) -return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + } rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);
@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }
Signature · 1 line
2.34.1
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were cc'ed on this. Adding Jens to the party
Quoted 1 line · Igor Opaniuk
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 14 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0
cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 9 lines
cmd/optee_rpmb.c +4
-2@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name, rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }
I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
Quoted 25 lines
cmd/optee_rpmb.c +6
-1rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char *name,struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)
cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }
ditto
Quoted 14 lines
cmd/optee_rpmb.c +3
-1rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; } --
Quoted 2 lines
2.34.1
Thanks
/Ilias
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
Hi Igor,
Quoted 27 lines · Ilias Apalodimas
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were cc'ed on this. Adding Jens to the party
Quoted 1 line · Igor Opaniuk
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 15 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name,
Quoted 12 lines
cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 8 lines
cmd/optee_rpmb.c +4
-2rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
Looking at it again, we do need tee != NULL, but I think it's cleaner to add something like
if (!dev)
return -ENODEV
to __tee_shm_add() instead.
Cheers
/Ilias
Quoted 58 lines
Quoted 26 lines
cmd/optee_rpmb.c +6
-1rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char *name,struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 14 lines
cmd/optee_rpmb.c +3
-1rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; } --Quoted 2 lines
2.34.1
Thanks
/Ilias
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
Hi Ilias,
Quoted 45 lines · Ilias Apalodimas
Hi Igor,
Quoted 4 lines · Ilias Apalodimas
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were
cc'ed
Quoted 23 lines
on this. Adding Jens to the party
Quoted 1 line · Igor Opaniuk
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 15 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name,Quoted 12 lines
cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 8 lines
cmd/optee_rpmb.c +4
-2rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
Quoted 6 lines
Looking at it again, we do need tee != NULL, but I think it's cleaner to add something like
if (!dev)
return-ENODEV
to__tee_shm_add()instead.
Do you mind if I address that in a separate patch series, as tbh I don't want to add additional patches to the existing series?
Quoted 63 lines
Cheers
/IliasQuoted 12 lines
Quoted 12 lines
cmd/optee_rpmb.c +3
-0rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;Quoted 3 lines
cmd/optee_rpmb.c static int write_persistent_value(const char
Quoted 3 lines
cmd/optee_rpmb.c +0
-0return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char*name,
Quoted 43 lines
Quoted 11 lines
cmd/optee_rpmb.c +3
-1struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 14 lines
cmd/optee_rpmb.c +3
-1rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; } --Quoted 2 lines
2.34.1
Thanks
/Ilias
Signature · 7 lines
Best regards - Atentamente - Meilleures salutations Igor Opaniuk mailto: [email protected] skype: igor.opanyuk https://www.linkedin.com/in/iopaniuk <http://ua.linkedin.com/in/iopaniuk>
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
Hi Igor,
Quoted 49 lines · Igor Opaniuk
Hi Ilias,
Quoted 45 lines · Ilias Apalodimas
Hi Igor,
Quoted 39 lines · Ilias Apalodimas
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were cc'ed on this. Adding Jens to the party
Quoted 1 line · Igor Opaniuk
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 14 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 9 lines
cmd/optee_rpmb.c +4
-2@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name, rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
Quoted 8 lines
Quoted 5 lines
Looking at it again, we do need tee != NULL, but I think it's cleaner to add something like
if (!dev)
return-ENODEV
to__tee_shm_add()instead.Do you mind if I address that in a separate patch series, as tbh I don't want to add additional patches to the existing series?
We still completely change the functionality. So, the patchset is not a 'tiny cleanup', it instead closes the session every time instead of keeping it open.
There are a few things going on, that aren't explained clearly in the commit message
- Why is this needed? Looking at the code it is an actual problem since sessions tied to the AVB uuid will remain open
- Is there any side effect by always closing the session? I don't mind merging it as is with a proper commit message, but I think the alternative is just easier.
Thanks
/Ilias
Quoted 75 lines
Quoted 63 lines
Cheers
/IliasQuoted 58 lines
Quoted 26 lines
cmd/optee_rpmb.c +6
-1rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char *name,struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 14 lines
cmd/optee_rpmb.c +3
-1rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; } --Quoted 2 lines
2.34.1
Thanks
/Ilias--
Best regards - Atentamente - Meilleures salutationsIgor Opaniuk
mailto: [email protected]
skype: igor.opanyuk
https://www.linkedin.com/in/iopaniuk
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
Ilias,
Quoted 66 lines · Ilias Apalodimas
Hi Igor,
Quoted 4 lines · Igor Opaniuk
Hi Ilias,
On Thu, Apr 4, 2024 at 10:54 AM Ilias Apalodimas <
Quoted 9 lines · [email protected]> wrote:
Quoted 9 lines
Hi Igor,
Quoted 4 lines · Ilias Apalodimas
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were
cc'ed
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
on this. Adding Jens to the party
Quoted 1 line · Igor Opaniuk
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of
why
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
this is needed
Quoted 6 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
cmd/optee_rpmb.c static int read_persistent_value(const char
Quoted 8 lines
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0@@ -87,8 +87,10 @@ static int read_persistent_value(const char*name,
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 9 lines
cmd/optee_rpmb.c +4
-2rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call
before
Quoted 10 lines
Quoted 7 lines
Quoted 1 line
opening the session?
Looking at it again, we do need tee != NULL, but I think it's cleaner to add something like
if (!dev)
return-ENODEV
to__tee_shm_add()instead.Do you mind if I address that in a separate patch series, as tbh I don't want to add additional patches to the existing series?
Quoted 11 lines
We still completely change the functionality. So, the patchset is not a 'tiny cleanup', it instead closes the session every time instead of keeping it open.
There are a few things going on, that aren't explained clearly in the commit message
- Why is this needed? Looking at the code it is an actual problem since sessions tied to the AVB uuid will remain open
- Is there any side effect by always closing the session? I don't mind merging it as is with a proper commit message, but I think the alternative is just easier.
I'll provide a bit more context here. The problem initially was in the TEE sandbox
driver implementation(drivers/tee/sandbox.c) and it's limitations, which doesn't
permit to have multiple simultaneous sessions with different TAs. This is what actually happened in this CI run [1], firstly "optee_rpmb" cmd was executed (and after execution we had one session open), and then "scp03", which also makes calls to OP-TEE, however it fails in sandbox_tee_open_session() because of this check:
if (state->ta) {
printf("A session is already open\n");
return -EBUSY;
}So basically I had two ways in mind to address that:
1. Close a session on each optee_rpmb cmd invocation.
I don't see any reason to keep this session open, as obviously there is no other mechanism (tbh, I don't know if DM calls ".remove" for active
devices) to close it automatically before handing over control to Linux kernel. As a result we might end up with some orphaned sessions registered in OP-TEE OS core (obvious resource leak).
2. Extend TEE sandbox driver, add support for multiple simultaneous sessions just to handle the case.
I've chosen the first approach, as IMO it was "kill two birds with one stone",
I could address resource leak in OP-TEE and bypass limitations of TEE sandbox driver.
[1] https://source.denx.de/u-boot/custodians/u-boot-tpm/-/jobs/792216
Quoted 81 lines
Thanks
/IliasQuoted 17 lines
Quoted 17 lines
Cheers
/IliasQuoted 12 lines
Quoted 12 lines
cmd/optee_rpmb.c +3
-0rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
cmd/optee_rpmb.c static int write_persistent_value(const char
Quoted 3 lines
cmd/optee_rpmb.c +0
-0return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char*name,
Quoted 55 lines
Quoted 35 lines
Quoted 35 lines
Quoted 11 lines
cmd/optee_rpmb.c +3
-1struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 11 lines
cmd/optee_rpmb.c +3
-0rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;Quoted 8 lines
Quoted 8 lines
Quoted 5 lines
cmd/optee_rpmb.c +0
-1return rc; } --2.34.1
Thanks
/Ilias--
Best regards - Atentamente - Meilleures salutationsIgor Opaniuk
mailto: [email protected]
skype: igor.opanyuk
https://www.linkedin.com/in/iopaniuk
Signature · 7 lines
Best regards - Atentamente - Meilleures salutations Igor Opaniuk mailto: [email protected] skype: igor.opanyuk https://www.linkedin.com/in/iopaniuk <http://ua.linkedin.com/in/iopaniuk>
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
Hi Igor,
Quoted 100 lines · Igor Opaniuk
Ilias,
Quoted 62 lines · Ilias Apalodimas
Hi Igor,
Quoted 49 lines · Igor Opaniuk
Hi Ilias,
Quoted 45 lines · Ilias Apalodimas
Hi Igor,
Quoted 39 lines · Ilias Apalodimas
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were cc'ed on this. Adding Jens to the party
Quoted 1 line · Igor Opaniuk
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 14 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 9 lines
cmd/optee_rpmb.c +4
-2@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name, rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
Quoted 8 lines
Quoted 5 lines
Looking at it again, we do need tee != NULL, but I think it's cleaner to add something like
if (!dev)
return-ENODEV
to__tee_shm_add()instead.Do you mind if I address that in a separate patch series, as tbh I don't want to add additional patches to the existing series?
Quoted 10 lines
We still completely change the functionality. So, the patchset is not a 'tiny cleanup', it instead closes the session every time instead of keeping it open.
There are a few things going on, that aren't explained clearly in the commit message
- Why is this needed? Looking at the code it is an actual problem since sessions tied to the AVB uuid will remain open
- Is there any side effect by always closing the session? I don't mind merging it as is with a proper commit message, but I think the alternative is just easier.I'll provide a bit more context here. The problem initially was in the TEE sandbox driver implementation(drivers/tee/sandbox.c) and it's limitations, which doesn't permit to have multiple simultaneous sessions with different TAs. This is what actually happened in this CI run [1], firstly "
optee_rpmb" cmd was executed (and after execution we had one session open), and then "scp03", which also makes calls to OP-TEE, however it fails
insandbox_tee_open_session()because of this check:if (state->ta) { printf("A session is already open\n"); return -EBUSY; }So basically I had two ways in mind to address that:
1. Close a session on eachoptee_rpmbcmd invocation.
I don't see any reason to keep this session open, as obviously there is no other mechanism (tbh, I don't know if DM calls ".remove" for active devices) to close it automatically before handing over control to Linux kernel. As a result we might end up with some orphaned sessions registered in OP-TEE OS core (obvious resource leak).
2. Extend TEE sandbox driver, add support for multiple simultaneous sessions just to handle the case.
Quoted 3 lines
I've chosen the first approach, as IMO it was "kill two birds with one stone", I could address resource leak in OP-TEE and bypass limitations of TEE sandbox driver.
Thanks, this helps.
All this needs to be in the commit message, instead of a mail thread. Can you send a new revision with the commit message amended?
Thanks
/Ilias
Quoted 94 lines
[1] https://source.denx.de/u-boot/custodians/u-boot-tpm/-/jobs/792216
Quoted 80 lines
Thanks
/IliasQuoted 75 lines
Quoted 63 lines
Cheers
/IliasQuoted 58 lines
Quoted 26 lines
cmd/optee_rpmb.c +6
-1rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char *name,struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 14 lines
cmd/optee_rpmb.c +3
-1rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; } --Quoted 2 lines
2.34.1
Thanks
/Ilias--
Best regards - Atentamente - Meilleures salutationsIgor Opaniuk
mailto: [email protected]
skype: igor.opanyuk
https://www.linkedin.com/in/iopaniuk--
Best regards - Atentamente - Meilleures salutationsIgor Opaniuk
mailto: [email protected]
skype: igor.opanyuk
https://www.linkedin.com/in/iopaniuk
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
In reply to Ilias ApalodimasHi Ilias,
Quoted 5 lines · Ilias Apalodimas
Hi Igor,
I was about to apply the series, but noticed neither me or Jens were cc'ed on this. Adding Jens to the party
I usually rely on automatic CC-list creation by patman. Looks like there is no
entry in MAINTAINERS for this specific file, that's why only Tom was added. I'll send a patch for that if you don't mind.
Quoted 21 lines · Igor Opaniuk
Quoted 1 line
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 15 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name,
Quoted 13 lines
cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 8 lines
cmd/optee_rpmb.c +4
-2rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
That's correct, but avb_ta_open_session() wrapper calls
both tee_find_device()/tee_open_session(), and tee_shm_alloc() expects valid tee device as param, which
is initialized by tee_find_device(). My intention was to address the only one particular issue
that was catched by CI and explained in [1] instead of doing overhaul of the whole optee_rpmb cmd implementation.
Quoted 49 lines
Quoted 25 lines
cmd/optee_rpmb.c +6
-1rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char *name,struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 11 lines
cmd/optee_rpmb.c +3
-0rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;
Quoted 9 lines
Quoted 5 lines
cmd/optee_rpmb.c +0
-1return rc; } --2.34.1
Thanks
/Ilias
[1]
https://lore.kernel.org/all/[email protected]/T/
Signature · 7 lines
Best regards - Atentamente - Meilleures salutations Igor Opaniuk mailto: [email protected] skype: igor.opanyuk https://www.linkedin.com/in/iopaniuk <http://ua.linkedin.com/in/iopaniuk>
Re: [PATCH v4 2/5] cmd: optee_rpmb: close tee session
HI Ilias
Quoted 7 lines · Igor Opaniuk
Hi Ilias,
Quoted 2 lines · Ilias Apalodimas
Hi Igor,
Quoted 8 lines
Quoted 3 lines
I was about to apply the series, but noticed neither me or Jens were cc'ed on this. Adding Jens to the party
I usually rely on automatic CC-list creation by patman. Looks like there is no
entry inMAINTAINERSfor this specific file, that's why only Tom was added. I'll send a patch for that if you don't mind.
I've added all orphaned tee-related files to MAINTAINERS in [1], so I patman should generate the CC list correctly next time.
[1]
https://lore.kernel.org/u-boot/[email protected]/T/#u
Quoted 118 lines
Quoted 23 lines · Igor Opaniuk
Quoted 1 line
Add calls for closing tee session after every read/write operation.
What the patch is pretty obvious, but I am missing an explanation of why this is needed
Quoted 15 lines
Signed-off-by Igor Opaniuk <[email protected]>---
(no changes since v1)
cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)cmd/optee_rpmb.c +0
-0@@ -87,8 +87,10 @@ static int read_persistent_value(const char *name,Quoted 13 lines
cmd/optee_rpmb.c static int read_persistent_value(const char *name, · line 93
Quoted 8 lines
cmd/optee_rpmb.c +4
-2rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }I don't think you need the session to be opened to allocate memory. So instead of of doing this, why don't we just move the alloc call before opening the session?
That's correct, but
avb_ta_open_session()wrapper calls
both tee_find_device()/tee_open_session(), andtee_shm_alloc()expects valid tee device as param, which
is initialized bytee_find_device(). My intention was to address the only one particular issue
that was catched by CI and explained in [1] instead of doing overhaul of the wholeoptee_rpmbcmd implementation.Quoted 50 lines
cmd/optee_rpmb.c static int write_persistent_value(const char
Quoted 14 lines
cmd/optee_rpmb.c +3
-0rc =tee_shm_alloc (tee, buffer_size,TEE_SHM_ALLOC , &shm_buf);@@ -125,6 +127,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;return rc; }@@ -139,17 +144,20 @@ static int write_persistent_value(const char*name,
Quoted 11 lines
cmd/optee_rpmb.c +3
-1struct tee_param param[2 ];size_t name_size =strlen (name) +1 ; +if (!value_size) +return -EINVAL ; +if (!tee) {if (avb_ta_open_session ())return -ENODEV ; } -if (!value_size)cmd/optee_rpmb.c static int write_persistent_value(const char *name, · line 160
Quoted 10 lines
cmd/optee_rpmb.c +4
-3-return -EINVAL ; rc =tee_shm_alloc (tee, name_size,TEE_SHM_ALLOC , &shm_name); -if (rc) -return -ENOMEM ; +if (rc) { + rc = -ENOMEM ; + goto close_session; + }ditto
Quoted 11 lines
cmd/optee_rpmb.c +3
-0rc =tee_shm_alloc (tee, value_size,TEE_SHM_ALLOC , &shm_buf);@@ -178,6 +186,9 @@ out:tee_shm_free (shm_buf); free_name:tee_shm_free (shm_name); +close_session: +tee_close_session (tee, session); + tee =NULL ;Quoted 9 lines
Quoted 5 lines
cmd/optee_rpmb.c +0
-1return rc; } --2.34.1
Thanks
/Ilias[1]
https://lore.kernel.org/all/[email protected]/T/--
Best regards - Atentamente - Meilleures salutationsIgor Opaniuk
mailto: [email protected]
skype: igor.opanyuk
https://www.linkedin.com/in/iopaniuk <http://ua.linkedin.com/in/iopaniuk>
Signature · 7 lines
Best regards - Atentamente - Meilleures salutations Igor Opaniuk mailto: [email protected] skype: igor.opanyuk https://www.linkedin.com/in/iopaniuk <http://ua.linkedin.com/in/iopaniuk>
PATCH v4 3/5 cmd: optee_rpmb: build cmd for sandbox
In reply to Igor OpaniukSupport CMD_OPTEE_RPMB for SANDBOX configurations. Test:
$ ./u-boot -d arch/sandbox/dts/test.dtb ... => optee_rpmb write_pvalue test_variable test_value Wrote 11 bytes => optee_rpmb read_pvalue test_variable 11 Read 11 bytes, value = test_value
---
(no changes since v2)
Changes in v2:
- Applied R-b and T-b tags
cmd/Kconfig | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
cmd/Kconfig +3 -1
@@ -1394,7 +1394,9 @@ config CMD_CLONE config CMD_OPTEE_RPMB bool "Enable read/write support on RPMB via OPTEE" - depends on SUPPORT_EMMC_RPMB && OPTEE + depends on (SUPPORT_EMMC_RPMB && OPTEE) || SANDBOX_TEE + default y if SANDBOX_TEE + select OPTEE_TA_AVB if SANDBOX_TEE help Enable the commands for reading, writing persistent named values in the Replay Protection Memory Block partition in eMMC by
Signature · 1 line
2.34.1
PATCH v4 4/5 test: py: add optee_rpmb tests
In reply to Igor OpaniukAdd read/write tests for optee_rpmb cmd.
---
(no changes since v1)
test/py/tests/test_optee_rpmb.py | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) create mode 100644 test/py/tests/test_optee_rpmb.py
test/py/tests/test_optee_rpmb.py new file +20 -0
@@ -0,0 +1,20 @@ +# SPDX-License-Identifier: GPL-2.0+ +# +# Tests for OP-TEE RPMB read/write support + +""" +This tests optee_rpmb cmd in U-Boot +""" + +import pytest +import u_boot_utils as util + [email protected]('cmd_optee_rpmb') +def test_optee_rpmb_read_write(u_boot_console): + """Test OP-TEE RPMB cmd read/write + """ + response = u_boot_console.run_command('optee_rpmb write_pvalue test_variable test_value') + assert response == 'Wrote 11 bytes' + + response = u_boot_console.run_command('optee_rpmb read_pvalue test_variable 11') + assert response == 'Read 11 bytes, value = test_value'
Signature · 1 line
2.34.1
PATCH v4 5/5 tee: remove common.h inclusion
In reply to Igor OpaniukThe usage of the common.h include file is deprecated [1], and has already been removed from several files.
Get rid of all inclusions in the "drivers/tee" directory, and replace it with required include files directly where needed.
[1] doc/develop/codingstyle.rst
---
(no changes since v3)
Changes in v3:
- Added calls for closing tee session after every read/write operation
- Added calls for closing tee session after every read/write operation
Changes in v2:
- Fixed chimp_optee.c:37:9: error: implicit declaration of function 'memset'
- Applied R-b and T-b tags
- Fixed chimp_optee.c:37:9: error: implicit declaration of function 'memset'
drivers/tee/broadcom/chimp_optee.c | 3 ++- drivers/tee/optee/core.c | 1 - drivers/tee/optee/i2c.c | 1 - drivers/tee/optee/rpmb.c | 1 - drivers/tee/optee/supplicant.c | 2 +- drivers/tee/sandbox.c | 2 +- drivers/tee/tee-uclass.c | 1 - 7 files changed, 4 insertions(+), 7 deletions(-)
drivers/tee/broadcom/chimp_optee.c +2 -1
@@ -3,9 +3,10 @@* Copyright 2020 Broadcom. */ -#include <common.h> #include <tee.h> #include <broadcom/chimp.h> +#include <linux/errno.h> +#include <string.h> #ifdef CONFIG_CHIMP_OPTEE
drivers/tee/optee/core.c +0 -1
@@ -3,7 +3,6 @@* Copyright (c) 2018-2020 Linaro Limited */ -#include <common.h> #include <cpu_func.h> #include <dm.h> #include <dm/device_compat.h>
drivers/tee/optee/i2c.c +0 -1
@@ -3,7 +3,6 @@* Copyright (c) 2020 Foundries.io Ltd */ -#include <common.h> #include <dm.h> #include <i2c.h> #include <tee.h>
drivers/tee/optee/rpmb.c +0 -1
@@ -3,7 +3,6 @@* Copyright (c) 2018 Linaro Limited */ -#include <common.h> #include <dm.h> #include <log.h> #include <tee.h>
drivers/tee/optee/supplicant.c +1 -1
@@ -3,10 +3,10 @@* Copyright (c) 2018, Linaro Limited */ -#include <common.h> #include <log.h> #include <malloc.h> #include <tee.h> +#include <linux/errno.h> #include <linux/types.h> #include "optee_msg.h"
drivers/tee/sandbox.c +1 -1
@@ -2,7 +2,7 @@/* * Copyright (C) 2018 Linaro Limited */ -#include <common.h> +#include <dm.h> #include <sandboxtee.h> #include <tee.h>
drivers/tee/tee-uclass.c +0 -1
@@ -5,7 +5,6 @@#define LOG_CATEGORY UCLASS_TEE -#include <common.h> #include <cpu_func.h> #include <dm.h> #include <log.h>
Signature · 1 line
2.34.1
Re: [PATCH v4 5/5] tee: remove common.h inclusion
I've already reviewed this one. [0]
I'll add my tag
[0] https://lore.kernel.org/u-boot/CAC_iWjLZcKEo6-0P_gkz5UOe7VanuNv1+=i3xcUTKVo5B1tpQg@mail.gmail.com/
Cheers
/Ilias
Quoted 127 lines · Igor Opaniuk
The usage of the
common.hinclude file is deprecated [1], and has already been removed from several files.
Get rid of all inclusions in the "drivers/tee" directory, and replace it with required include files directly where needed.[1]
doc/develop/codingstyle.rstSigned-off-by Igor Opaniuk <[email protected]>---
(no changes since v3)
Changes in v3:
- Added calls for closing tee session after every read/write operation
- Added calls for closing tee session after every read/write operationChanges in v2:
- Fixed chimp_optee.c:37:9: error: implicit declaration of function 'memset'
- Applied R-b and T-b tags
- Fixed chimp_optee.c:37:9: error: implicit declaration of function 'memset'drivers/tee/broadcom/chimp_optee.c | 3 ++- drivers/tee/optee/core.c | 1 - drivers/tee/optee/i2c.c | 1 - drivers/tee/optee/rpmb.c | 1 - drivers/tee/optee/supplicant.c | 2 +- drivers/tee/sandbox.c | 2 +- drivers/tee/tee-uclass.c | 1 - 7 files changed, 4 insertions(+), 7 deletions(-)drivers/tee/broadcom/chimp_optee.c +2
-1@@ -3,9 +3,10 @@* Copyright 2020 Broadcom. */ -#include <common.h> #include <tee.h> #include <broadcom/chimp.h> +#include <linux/errno.h> +#include <string.h> #ifdef CONFIG_CHIMP_OPTEE drivers/tee/optee/core.c +0
-1@@ -3,7 +3,6 @@* Copyright (c) 2018-2020 Linaro Limited */ -#include <common.h> #include <cpu_func.h> #include <dm.h> #include <dm/device_compat.h> drivers/tee/optee/i2c.c +0
-1@@ -3,7 +3,6 @@* Copyright (c) 2020 Foundries.io Ltd */ -#include <common.h> #include <dm.h> #include <i2c.h> #include <tee.h> drivers/tee/optee/rpmb.c +0
-1@@ -3,7 +3,6 @@* Copyright (c) 2018 Linaro Limited */ -#include <common.h> #include <dm.h> #include <log.h> #include <tee.h> drivers/tee/optee/supplicant.c +1
-1@@ -3,10 +3,10 @@* Copyright (c) 2018, Linaro Limited */ -#include <common.h> #include <log.h> #include <malloc.h> #include <tee.h> +#include <linux/errno.h> #include <linux/types.h> #include "optee_msg.h" drivers/tee/sandbox.c +1
-1@@ -2,7 +2,7 @@/* * Copyright (C) 2018 Linaro Limited */ -#include <common.h> +#include <dm.h> #include <sandboxtee.h> #include <tee.h> drivers/tee/tee-uclass.c +0
-2@@ -5,7 +5,6 @@#define LOG_CATEGORY UCLASS_TEE -#include <common.h> #include <cpu_func.h> #include <dm.h> #include <log.h> --2.34.1
Thanks Igor
I'll pick this up
Quoted 44 lines · Igor Opaniuk
- Address some spelling errors and typos
- SupportCMD_OPTEE_RPMBfor SANDBOX configurations and add python tests
- Removecommon.hinclusion fordrivers/tee
- Add calls for closing tee session after every read/write operation.CI build:
[1] https://dev.azure.com/igoropaniuk/u-boot/_build/results?buildId=31&view=resultsChanges in v4:
- Rebased on the latest master and excluded "tee: sandbox: fix spelling errors", as it was merged already by Heinrich SchuchardtChanges in v3:
- Added calls for closing tee session after every read/write operationChanges in v2:
- Fixed chimp_optee.c:37:9: error: implicit declaration of function 'memset'
- Applied R-b and T-b tagsIgor Opaniuk (5):
tee: optee: fix description inKconfig
cmd:optee_rpmb: close tee session
cmd:optee_rpmb: build cmd for sandbox
test: py: addoptee_rpmbtests
tee: removecommon.hinclusioncmd/Kconfig | 4 +++- cmd/optee_rpmb.c | 23 +++++++++++++++++------ drivers/tee/broadcom/chimp_optee.c | 3 ++- drivers/tee/optee/Kconfig | 2 +- drivers/tee/optee/core.c | 1 - drivers/tee/optee/i2c.c | 1 - drivers/tee/optee/rpmb.c | 1 - drivers/tee/optee/supplicant.c | 2 +- drivers/tee/sandbox.c | 2 +- drivers/tee/tee-uclass.c | 1 - test/py/tests/test_optee_rpmb.py | 20 ++++++++++++++++++++ 11 files changed, 45 insertions(+), 15 deletions(-) create mode 100644 test/py/tests/test_optee_rpmb.py--
2.34.1