Message ID | 20240404081352.3224643-3-igor.opaniuk@gmail.com |
---|---|
State | Superseded |
Delegated to: | Ilias Apalodimas |
Headers | show |
Series | TEE: minor cleanup | expand |
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 On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: > 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 > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> > --- > > (no changes since v1) > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ > 1 file changed, 17 insertions(+), 6 deletions(-) > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c > index e0e44bbed04..b3cafd92410 100644 > --- a/cmd/optee_rpmb.c > +++ b/cmd/optee_rpmb.c > @@ -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? > > 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; > + } ditto > > 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; > } > -- > 2.34.1 > Thanks /Ilias
Hi Igor, On Thu, 4 Apr 2024 at 11:40, Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: > > 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 > > On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: > > 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 > > > > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> > > --- > > > > (no changes since v1) > > > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ > > 1 file changed, 17 insertions(+), 6 deletions(-) > > > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c > > index e0e44bbed04..b3cafd92410 100644 > > --- a/cmd/optee_rpmb.c > > +++ b/cmd/optee_rpmb.c > > @@ -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? 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 > > > > > 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; > > + } > > ditto > > > > > 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; > > } > > -- > > 2.34.1 > > > > Thanks > /Ilias
Hi Ilias, On Thu, Apr 4, 2024 at 10:40 AM Ilias Apalodimas < ilias.apalodimas@linaro.org> wrote: > 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. > On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: > > 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 > > > > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> > > --- > > > > (no changes since v1) > > > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ > > 1 file changed, 17 insertions(+), 6 deletions(-) > > > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c > > index e0e44bbed04..b3cafd92410 100644 > > --- a/cmd/optee_rpmb.c > > +++ b/cmd/optee_rpmb.c > > @@ -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? > 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. > > > > 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; > > + } > > ditto > > > > > 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; > > } > > -- > > 2.34.1 > > > > Thanks > /Ilias > [1] https://lore.kernel.org/all/20240302220108.720637-5-igor.opaniuk@gmail.com/T/
Hi Ilias, On Thu, Apr 4, 2024 at 10:54 AM Ilias Apalodimas < ilias.apalodimas@linaro.org> wrote: > Hi Igor, > > On Thu, 4 Apr 2024 at 11:40, Ilias Apalodimas > <ilias.apalodimas@linaro.org> wrote: > > > > 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 > > > > On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: > > > 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 > > > > > > > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> > > > --- > > > > > > (no changes since v1) > > > > > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ > > > 1 file changed, 17 insertions(+), 6 deletions(-) > > > > > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c > > > index e0e44bbed04..b3cafd92410 100644 > > > --- a/cmd/optee_rpmb.c > > > +++ b/cmd/optee_rpmb.c > > > @@ -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? > > 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? > > Cheers > /Ilias > > > > > > > > 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; > > > + } > > > > ditto > > > > > > > > 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; > > > } > > > -- > > > 2.34.1 > > > > > > > Thanks > > /Ilias >
HI Ilias On Thu, Apr 4, 2024 at 12:15 PM Igor Opaniuk <igor.opaniuk@gmail.com> wrote: > Hi Ilias, > > On Thu, Apr 4, 2024 at 10:40 AM Ilias Apalodimas < > ilias.apalodimas@linaro.org> wrote: > >> 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. > 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/20240404105248.3241506-1-igor.opaniuk@gmail.com/T/#u > > >> On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: >> > 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 >> >> > >> > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> >> > --- >> > >> > (no changes since v1) >> > >> > cmd/optee_rpmb.c | 23 +++++++++++++++++------ >> > 1 file changed, 17 insertions(+), 6 deletions(-) >> > >> > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c >> > index e0e44bbed04..b3cafd92410 100644 >> > --- a/cmd/optee_rpmb.c >> > +++ b/cmd/optee_rpmb.c >> > @@ -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? >> > 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. > > >> > >> > 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; >> > + } >> >> ditto >> >> > >> > 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; >> > } >> > -- >> > 2.34.1 >> > >> >> Thanks >> /Ilias >> > > [1] > https://lore.kernel.org/all/20240302220108.720637-5-igor.opaniuk@gmail.com/T/ > > -- > Best regards - Atentamente - Meilleures salutations > > Igor Opaniuk > > mailto: igor.opaniuk@gmail.com > skype: igor.opanyuk > https://www.linkedin.com/in/iopaniuk <http://ua.linkedin.com/in/iopaniuk> >
Hi Igor, On Thu, 4 Apr 2024 at 13:18, Igor Opaniuk <igor.opaniuk@gmail.com> wrote: > > Hi Ilias, > > On Thu, Apr 4, 2024 at 10:54 AM Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: >> >> Hi Igor, >> >> On Thu, 4 Apr 2024 at 11:40, Ilias Apalodimas >> <ilias.apalodimas@linaro.org> wrote: >> > >> > 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 >> > >> > On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: >> > > 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 >> > >> > > >> > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> >> > > --- >> > > >> > > (no changes since v1) >> > > >> > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ >> > > 1 file changed, 17 insertions(+), 6 deletions(-) >> > > >> > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c >> > > index e0e44bbed04..b3cafd92410 100644 >> > > --- a/cmd/optee_rpmb.c >> > > +++ b/cmd/optee_rpmb.c >> > > @@ -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? >> >> 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 > >> >> >> Cheers >> /Ilias >> > >> > > >> > > 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; >> > > + } >> > >> > ditto >> > >> > > >> > > 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; >> > > } >> > > -- >> > > 2.34.1 >> > > >> > >> > Thanks >> > /Ilias > > > > > -- > Best regards - Atentamente - Meilleures salutations > > Igor Opaniuk > > mailto: igor.opaniuk@gmail.com > skype: igor.opanyuk > https://www.linkedin.com/in/iopaniuk
Ilias, On Thu, Apr 4, 2024 at 1:00 PM Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: > Hi Igor, > > On Thu, 4 Apr 2024 at 13:18, Igor Opaniuk <igor.opaniuk@gmail.com> wrote: > > > > Hi Ilias, > > > > On Thu, Apr 4, 2024 at 10:54 AM Ilias Apalodimas < > ilias.apalodimas@linaro.org> wrote: > >> > >> Hi Igor, > >> > >> On Thu, 4 Apr 2024 at 11:40, Ilias Apalodimas > >> <ilias.apalodimas@linaro.org> wrote: > >> > > >> > 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 > >> > > >> > On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: > >> > > 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 > >> > > >> > > > >> > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> > >> > > --- > >> > > > >> > > (no changes since v1) > >> > > > >> > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ > >> > > 1 file changed, 17 insertions(+), 6 deletions(-) > >> > > > >> > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c > >> > > index e0e44bbed04..b3cafd92410 100644 > >> > > --- a/cmd/optee_rpmb.c > >> > > +++ b/cmd/optee_rpmb.c > >> > > @@ -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? > >> > >> 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. > 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 > > Thanks > /Ilias > > > > >> > >> > >> Cheers > >> /Ilias > >> > > >> > > > >> > > 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; > >> > > + } > >> > > >> > ditto > >> > > >> > > > >> > > 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; > >> > > } > >> > > -- > >> > > 2.34.1 > >> > > > >> > > >> > Thanks > >> > /Ilias > > > > > > > > > > -- > > Best regards - Atentamente - Meilleures salutations > > > > Igor Opaniuk > > > > mailto: igor.opaniuk@gmail.com > > skype: igor.opanyuk > > https://www.linkedin.com/in/iopaniuk >
Hi Igor, On Thu, 4 Apr 2024 at 14:40, Igor Opaniuk <igor.opaniuk@gmail.com> wrote: > > Ilias, > > On Thu, Apr 4, 2024 at 1:00 PM Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: >> >> Hi Igor, >> >> On Thu, 4 Apr 2024 at 13:18, Igor Opaniuk <igor.opaniuk@gmail.com> wrote: >> > >> > Hi Ilias, >> > >> > On Thu, Apr 4, 2024 at 10:54 AM Ilias Apalodimas <ilias.apalodimas@linaro.org> wrote: >> >> >> >> Hi Igor, >> >> >> >> On Thu, 4 Apr 2024 at 11:40, Ilias Apalodimas >> >> <ilias.apalodimas@linaro.org> wrote: >> >> > >> >> > 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 >> >> > >> >> > On Thu, Apr 04, 2024 at 10:13:49AM +0200, Igor Opaniuk wrote: >> >> > > 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 >> >> > >> >> > > >> >> > > Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> >> >> > > --- >> >> > > >> >> > > (no changes since v1) >> >> > > >> >> > > cmd/optee_rpmb.c | 23 +++++++++++++++++------ >> >> > > 1 file changed, 17 insertions(+), 6 deletions(-) >> >> > > >> >> > > diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c >> >> > > index e0e44bbed04..b3cafd92410 100644 >> >> > > --- a/cmd/optee_rpmb.c >> >> > > +++ b/cmd/optee_rpmb.c >> >> > > @@ -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? >> >> >> >> 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. > > 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. 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 > > [1] https://source.denx.de/u-boot/custodians/u-boot-tpm/-/jobs/792216 > >> >> >> Thanks >> /Ilias >> >> > >> >> >> >> >> >> Cheers >> >> /Ilias >> >> > >> >> > > >> >> > > 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; >> >> > > + } >> >> > >> >> > ditto >> >> > >> >> > > >> >> > > 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; >> >> > > } >> >> > > -- >> >> > > 2.34.1 >> >> > > >> >> > >> >> > Thanks >> >> > /Ilias >> > >> > >> > >> > >> > -- >> > Best regards - Atentamente - Meilleures salutations >> > >> > Igor Opaniuk >> > >> > mailto: igor.opaniuk@gmail.com >> > skype: igor.opanyuk >> > https://www.linkedin.com/in/iopaniuk > > > > -- > Best regards - Atentamente - Meilleures salutations > > Igor Opaniuk > > mailto: igor.opaniuk@gmail.com > skype: igor.opanyuk > https://www.linkedin.com/in/iopaniuk
diff --git a/cmd/optee_rpmb.c b/cmd/optee_rpmb.c index e0e44bbed04..b3cafd92410 100644 --- a/cmd/optee_rpmb.c +++ b/cmd/optee_rpmb.c @@ -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; }
Add calls for closing tee session after every read/write operation. Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> --- (no changes since v1) cmd/optee_rpmb.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-)