| Message ID | 20180418073259.18645-1-sven.eckelmann@openmesh.com (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Delegated to: | Andy Gross |
| Headers | show |
Hi, On 4/18/2018 1:02 PM, Sven Eckelmann wrote: > The QSEE (trustzone) is started on IPQ4019 before Linux is started. > According to QCA, it is placed in in the the memory region > 0x87e80000-0x88000000 and must not be accessed directly. There is an > additional memory region 0x87e00000-0x87E80000 smem which which can be used > for communication with the TZ. The driver for the latter is not yet ready > but it is still not allowed to use this memory region like any other > memory region. > > Not reserving this memory region either leads to kernel crashes, kernel > hangs (often during the boot) or bus errors for userspace programs. The > latter happens when a program is using a memory region which is mapped to > these physical memory regions. > > [ 571.758058] Unhandled fault: imprecise external abort (0xc06) at 0x01715ff8 > [ 571.758099] pgd = cebec000 > [ 571.763826] [01715ff8] *pgd=8e7fa835, *pte=87e7f75f, *ppte=87e7fc7f > Bus error > > Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com> > --- Acked-by: Sricharan R <sricharan@codeaurora.org> Regards, Sricharan > Cc: Sricharan Ramabadhran <srichara@qti.qualcomm.com> > Cc: Senthilkumar N L <snlakshm@qti.qualcomm.com> > > There are additional memory regions which have to be initialized first by > Linux. So they are currently not used. We were told by QCA that the > features QSDK uses them for are: > > * crash dump feature > - a couple of regions used when 'qca,scm_restart_reason' dt node has the > value 'dload_status' not set to 1 > + apps_bl <0x87000000 0x400000> > + sbl <0x87400000 0x100000> > + cnss_debug <0x87400000 0x100000> > + cpu_context_dump <0x87b00000 0x080000> > - required driver not available in Linux > - safe to remove > * QSEE app execution > - region tz_apps <0x87b80000 0x280000> > - required driver not available in Linux > - safe to remove > * communication with TZ/QSEE > - region smem <0x87b80000 0x280000> > - driver changes not yet upstreamed > - must not be removed because any access can crash kernel/program > * trustzone (QSEE) private memory > - region tz <0x87e80000 0x180000> > - must not be removed because any access can crash kernel/program > > The problem with the missing regions was reported in 2016 [1]. So maybe > this change qualifies for a stable@vger.kernel.org submission. > > [1] https://www.spinics.net/lists/linux-arm-msm/msg21536.html > --- > arch/arm/boot/dts/qcom-ipq4019.dtsi | 16 ++++++++++++++++ > 1 file changed, 16 insertions(+) > > diff --git a/arch/arm/boot/dts/qcom-ipq4019.dtsi b/arch/arm/boot/dts/qcom-ipq4019.dtsi > index 10d112a4078e..b25daf368012 100644 > --- a/arch/arm/boot/dts/qcom-ipq4019.dtsi > +++ b/arch/arm/boot/dts/qcom-ipq4019.dtsi > @@ -23,6 +23,22 @@ > compatible = "qcom,ipq4019"; > interrupt-parent = <&intc>; > > + reserved-memory { > + #address-cells = <0x1>; > + #size-cells = <0x1>; > + ranges; > + > + smem_region: smem@87e00000 { > + reg = <0x87e00000 0x080000>; > + no-map; > + }; > + > + tz@87e80000 { > + reg = <0x87e80000 0x180000>; > + no-map; > + }; > + }; > + > aliases { > spi0 = &spi_0; > i2c0 = &i2c_0; >
diff --git a/arch/arm/boot/dts/qcom-ipq4019.dtsi b/arch/arm/boot/dts/qcom-ipq4019.dtsi index 10d112a4078e..b25daf368012 100644 --- a/arch/arm/boot/dts/qcom-ipq4019.dtsi +++ b/arch/arm/boot/dts/qcom-ipq4019.dtsi @@ -23,6 +23,22 @@ compatible = "qcom,ipq4019"; interrupt-parent = <&intc>; + reserved-memory { + #address-cells = <0x1>; + #size-cells = <0x1>; + ranges; + + smem_region: smem@87e00000 { + reg = <0x87e00000 0x080000>; + no-map; + }; + + tz@87e80000 { + reg = <0x87e80000 0x180000>; + no-map; + }; + }; + aliases { spi0 = &spi_0; i2c0 = &i2c_0;
The QSEE (trustzone) is started on IPQ4019 before Linux is started. According to QCA, it is placed in in the the memory region 0x87e80000-0x88000000 and must not be accessed directly. There is an additional memory region 0x87e00000-0x87E80000 smem which which can be used for communication with the TZ. The driver for the latter is not yet ready but it is still not allowed to use this memory region like any other memory region. Not reserving this memory region either leads to kernel crashes, kernel hangs (often during the boot) or bus errors for userspace programs. The latter happens when a program is using a memory region which is mapped to these physical memory regions. [ 571.758058] Unhandled fault: imprecise external abort (0xc06) at 0x01715ff8 [ 571.758099] pgd = cebec000 [ 571.763826] [01715ff8] *pgd=8e7fa835, *pte=87e7f75f, *ppte=87e7fc7f Bus error Signed-off-by: Sven Eckelmann <sven.eckelmann@openmesh.com> --- Cc: Sricharan Ramabadhran <srichara@qti.qualcomm.com> Cc: Senthilkumar N L <snlakshm@qti.qualcomm.com> There are additional memory regions which have to be initialized first by Linux. So they are currently not used. We were told by QCA that the features QSDK uses them for are: * crash dump feature - a couple of regions used when 'qca,scm_restart_reason' dt node has the value 'dload_status' not set to 1 + apps_bl <0x87000000 0x400000> + sbl <0x87400000 0x100000> + cnss_debug <0x87400000 0x100000> + cpu_context_dump <0x87b00000 0x080000> - required driver not available in Linux - safe to remove * QSEE app execution - region tz_apps <0x87b80000 0x280000> - required driver not available in Linux - safe to remove * communication with TZ/QSEE - region smem <0x87b80000 0x280000> - driver changes not yet upstreamed - must not be removed because any access can crash kernel/program * trustzone (QSEE) private memory - region tz <0x87e80000 0x180000> - must not be removed because any access can crash kernel/program The problem with the missing regions was reported in 2016 [1]. So maybe this change qualifies for a stable@vger.kernel.org submission. [1] https://www.spinics.net/lists/linux-arm-msm/msg21536.html --- arch/arm/boot/dts/qcom-ipq4019.dtsi | 16 ++++++++++++++++ 1 file changed, 16 insertions(+)