MPU access violation in SysCtlClockFreqSet when changing from Tivaware 2.1.1 to 2.1.3

Per E

Other Parts Discussed in Thread: TM4C1294NCPDT

Hi,

Since upgrading Tivaware to 2.1.3 I have seen a strange and very rare crash that was not present earlier. It happens during a call to SysCtlClockFreqSet (this function was rewritten in Tivaware 2.1.3):

SysCtlClockFreqSet(SYSCTL_OSC_INT |
                   SYSCTL_USE_PLL |
                   SYSCTL_CFG_VCO_480,
                   120000000);

Our register dump looks like this:

[!] MPU access violation
Data access at 0xF8D4FCF5
FAULTSTAT: 0x00000082
R0  = 0x00F42400
R1  = 0x07270E00
R2  = 0x0000000C
R3  = 0x00000000
R12 = 0xF8D4FCF1
LR  = 0x0005C45F
PC  = 0x00008FF8
PSR = 0x21000000

We jump between two execution modes, the other being

SysCtlClockFreqSet(SYSCTL_XTAL_25MHZ |
                   SYSCTL_OSC_MAIN |
                   SYSCTL_USE_PLL |
                   SYSCTL_CFG_VCO_480,
                   120000000);

Are these valid usages of SysCtlClockFreqSet? I am thinking of downgrading to 2.1.1 again but there are some other fixes in the new version that are good for us.

Kind regards Per

over 8 years ago

0 Per E over 8 years ago

Intellectual 490 points

To clarify, we are using TM4C1294NCPDT

0 Amit Ashara over 8 years ago

TI__Guru**** 244400 points

Hello Per

The configuration listed are valid configurations. You mean to say that when the execution moves from the 1st to 2nd there is a crash?

0 Per E over 8 years ago in reply to Amit Ashara

Intellectual 490 points

Yes, this was what I meant. However, we have only seen it once since the switch (and we have thorough regression tests) so I'm closing the post. *Knock on wood*

0 Per E over 8 years ago in reply to Amit Ashara

Intellectual 490 points

Maybe it could be related to e2e.ti.com/.../539105 ?

Was this bug introduced recently? I know SysCtlClockFreqSet was rewritten in one of the latest releases.

0 Amit Ashara over 8 years ago in reply to Per E

TI__Guru**** 244400 points

Hello per E

No this was not a bug but the correct operation of waiting for the MOSC to Power Up was not done.

0 Per E over 8 years ago in reply to Amit Ashara

Intellectual 490 points

I'm sorry Amit, I don't follow you ;) You say the function does not wait for the MOSC to power up but it is not classified as a bug?

0 Amit Ashara over 8 years ago in reply to Per E

TI__Guru**** 244400 points

Hello Per

Yes, that is correct. Technically it can be interpreted as a bug in the software and not a design limitation/error.

0 Per E over 8 years ago in reply to Amit Ashara

Intellectual 490 points

OK, let me rephrase.

If I interpret it as a bug (the function is/may under some circumstances be compromised the way TivaWare is written in combination with the silicon design), was this bug introduced in the recent Tivaware changes?

Kind regards Per

0 Amit Ashara over 8 years ago in reply to Per E

TI__Guru**** 244400 points

Hello Per

It was not there in the first place. I traced the code back to the first TivaWare release for TM4C129x device and the MOSCPU bit was not being checked.

Arm-based microcontrollers

Arm-based microcontrollers forum

MPU access violation in SysCtlClockFreqSet when changing from Tivaware 2.1.1 to 2.1.3