Message ID | 20240730013834.41840-2-mailingradian@gmail.com (mailing list archive) |
---|---|
State | Accepted |
Commit | 8342009efa2a5e75dce56173d7de026bcc6666d8 |
Headers | show |
Series | firmware: qcom: tzmem: disable sdm670 platform | expand |
On Tue, Jul 30, 2024 at 3:38 AM Richard Acayan <mailingradian@gmail.com> wrote: > > The Pixel 3a returns 4291821499 (-3145797 or 0xFFCFFFBB) when attempting > to load the GPU firmware if tzmem is allowed. Disable it on SDM670 so > the GPU can successfully probe. > > Signed-off-by: Richard Acayan <mailingradian@gmail.com> > --- > drivers/firmware/qcom/qcom_tzmem.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/firmware/qcom/qcom_tzmem.c b/drivers/firmware/qcom/qcom_tzmem.c > index 17948cfc82e7..5767ef210036 100644 > --- a/drivers/firmware/qcom/qcom_tzmem.c > +++ b/drivers/firmware/qcom/qcom_tzmem.c > @@ -78,6 +78,7 @@ static bool qcom_tzmem_using_shm_bridge; > /* List of machines that are known to not support SHM bridge correctly. */ > static const char *const qcom_tzmem_blacklist[] = { > "qcom,sc8180x", > + "qcom,sdm670", /* failure in GPU firmware loading */ > "qcom,sdm845", /* reset in rmtfs memory assignment */ > "qcom,sm8150", /* reset in rmtfs memory assignment */ > NULL > -- > 2.45.2 > > Ugh... As the list is growing post mainline merge I'm wondering whether we should make the generic allocator the default in defconfig and SHM Bridge an opt-in option? Bartosz
On Tue, Jul 30, 2024 at 12:27:44PM GMT, Bartosz Golaszewski wrote: > On Tue, Jul 30, 2024 at 3:38 AM Richard Acayan <mailingradian@gmail.com> wrote: > > > > The Pixel 3a returns 4291821499 (-3145797 or 0xFFCFFFBB) when attempting > > to load the GPU firmware if tzmem is allowed. Disable it on SDM670 so > > the GPU can successfully probe. > > > > Signed-off-by: Richard Acayan <mailingradian@gmail.com> > > --- > > drivers/firmware/qcom/qcom_tzmem.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/drivers/firmware/qcom/qcom_tzmem.c b/drivers/firmware/qcom/qcom_tzmem.c > > index 17948cfc82e7..5767ef210036 100644 > > --- a/drivers/firmware/qcom/qcom_tzmem.c > > +++ b/drivers/firmware/qcom/qcom_tzmem.c > > @@ -78,6 +78,7 @@ static bool qcom_tzmem_using_shm_bridge; > > /* List of machines that are known to not support SHM bridge correctly. */ > > static const char *const qcom_tzmem_blacklist[] = { > > "qcom,sc8180x", > > + "qcom,sdm670", /* failure in GPU firmware loading */ > > "qcom,sdm845", /* reset in rmtfs memory assignment */ > > "qcom,sm8150", /* reset in rmtfs memory assignment */ > > NULL > > -- > > 2.45.2 > > > > > > Ugh... As the list is growing post mainline merge I'm wondering > whether we should make the generic allocator the default in defconfig > and SHM Bridge an opt-in option? > Can you confirm that we still need this, and the other entries in this list? Thanks, Bjorn
On Thu, Aug 15, 2024 at 11:58 PM Bjorn Andersson <andersson@kernel.org> wrote: > > On Tue, Jul 30, 2024 at 12:27:44PM GMT, Bartosz Golaszewski wrote: > > On Tue, Jul 30, 2024 at 3:38 AM Richard Acayan <mailingradian@gmail.com> wrote: > > > > > > The Pixel 3a returns 4291821499 (-3145797 or 0xFFCFFFBB) when attempting > > > to load the GPU firmware if tzmem is allowed. Disable it on SDM670 so > > > the GPU can successfully probe. > > > > > > Signed-off-by: Richard Acayan <mailingradian@gmail.com> > > > --- > > > drivers/firmware/qcom/qcom_tzmem.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/drivers/firmware/qcom/qcom_tzmem.c b/drivers/firmware/qcom/qcom_tzmem.c > > > index 17948cfc82e7..5767ef210036 100644 > > > --- a/drivers/firmware/qcom/qcom_tzmem.c > > > +++ b/drivers/firmware/qcom/qcom_tzmem.c > > > @@ -78,6 +78,7 @@ static bool qcom_tzmem_using_shm_bridge; > > > /* List of machines that are known to not support SHM bridge correctly. */ > > > static const char *const qcom_tzmem_blacklist[] = { > > > "qcom,sc8180x", > > > + "qcom,sdm670", /* failure in GPU firmware loading */ > > > "qcom,sdm845", /* reset in rmtfs memory assignment */ > > > "qcom,sm8150", /* reset in rmtfs memory assignment */ > > > NULL > > > -- > > > 2.45.2 > > > > > > > > > > Ugh... As the list is growing post mainline merge I'm wondering > > whether we should make the generic allocator the default in defconfig > > and SHM Bridge an opt-in option? > > > > Can you confirm that we still need this, and the other entries in this > list? > > Thanks, > Bjorn Yes, the fix you picked up was for a different bug. I'm in touch with Qualcomm about debugging the firmware here but it's not going to be immediate so let's apply this for now. Acked-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Bart
On Mon, 29 Jul 2024 21:38:35 -0400, Richard Acayan wrote: > The Pixel 3a returns 4291821499 (-3145797 or 0xFFCFFFBB) when attempting > to load the GPU firmware if tzmem is allowed. Disable it on SDM670 so > the GPU can successfully probe. > > Applied, thanks! [1/1] firmware: qcom: tzmem: disable sdm670 platform commit: 8342009efa2a5e75dce56173d7de026bcc6666d8 Best regards,
diff --git a/drivers/firmware/qcom/qcom_tzmem.c b/drivers/firmware/qcom/qcom_tzmem.c index 17948cfc82e7..5767ef210036 100644 --- a/drivers/firmware/qcom/qcom_tzmem.c +++ b/drivers/firmware/qcom/qcom_tzmem.c @@ -78,6 +78,7 @@ static bool qcom_tzmem_using_shm_bridge; /* List of machines that are known to not support SHM bridge correctly. */ static const char *const qcom_tzmem_blacklist[] = { "qcom,sc8180x", + "qcom,sdm670", /* failure in GPU firmware loading */ "qcom,sdm845", /* reset in rmtfs memory assignment */ "qcom,sm8150", /* reset in rmtfs memory assignment */ NULL
The Pixel 3a returns 4291821499 (-3145797 or 0xFFCFFFBB) when attempting to load the GPU firmware if tzmem is allowed. Disable it on SDM670 so the GPU can successfully probe. Signed-off-by: Richard Acayan <mailingradian@gmail.com> --- drivers/firmware/qcom/qcom_tzmem.c | 1 + 1 file changed, 1 insertion(+)