topic MMU running problem in codewarrior 10.6 environment in CodeWarrior Development Tools

MMU running problem in codewarrior 10.6 environment

pengjianxin — Wed, 28 Nov 2018 10:03:51 GMT

hello：

A project （MCU type: MPC5644A） can run normally in codewarrior 2.10 . But in the codewarrior 10.6 environment the program crashes. the program only have one function （MmuSetRam256K） which is about MMU initial. when the program run the instruction " tlbwe " （Write entry defined in MAS0 (Entry0 here) to MMU TLB），the program dies both in FLASH and RAM mode.

the program is in the attachment，look forward to your reply, thanks.

Re: MMU running problem in codewarrior 10.6 environment

pengjianxin — Wed, 28 Nov 2018 10:07:13 GMT

looking forward to your response,My E@mal is" flesher@126.com ", thank you.

Re: MMU running problem in codewarrior 10.6 environment

martin_kovar — Thu, 06 Dec 2018 12:43:38 GMT

Hello,

I am able to reproduce this problem on my side, unfortunately, I have no idea, why the program does not work correct. I checked with Lauterbach debugger and there is no problem with functionality.

Only possibility for me is create bug ticket and wait, if this issue will be fixed in future.

I am sorry for any inconvenience.

Regards,

Martin

Re: MMU running problem in codewarrior 10.6 environment

takaoyamada — Fri, 07 Dec 2018 20:21:38 GMT

Greetings,

Try writing to MMU entry that is not zero. You are most likely replacing a previous entry that covered SRAM and/or FLASH and hence debugging fails. E.g. try this

e_lis r3, 0x1003 // Select TLB1 Entry 1, define R/W replacement control
ori r3, r3, 0x0000
mtMAS0 r3 // Load MAS0 with 0x1003 0000 for TLB Entry1

Within Codewarrior 2.10, the MMU table is initialized differently using a reset macro script from PEmicro. Within this script, there are multiple entries in the MMU that closely resembles how the BAF initializes the chip. The zeroth entry is Peripheral B and this is why FLASH and SRAM were kept intact and debugging behaved normally. Most likely Lauterbach is similarly setting up the MMU Table in this fashion and hence why there was no issue there as well. However in Codewarrior 10.x and 11.x the reset scripts are handled using NXP's TCL scripts that writes only one entry in the MMU Table at entry 0:

# Setup MMU entry 0 for entire 32-bit address space: base address = 0x0000_0000
# 4GB Memory Space, Not Guarded, Cache inhibited, All Access
if {$booke_vle == "vle"} {
# VLE page
reg ${TLB1_GROUP}MMU_CAM0 = 0xB0000008FE0800000000000000000001
} else {
# BOOKE page
reg ${TLB1_GROUP}MMU_CAM0 = 0xB0000008FC0800000000000000000001
}

This one entry is set to cover the entire memory map. Your entry is replacing this single MMU table entry and therefore you lose access to the flash and SRAM located elsewhere. This is why reading the current MMU table entries before writing to it is probably good practice.

Note, S32 Power IDE does use PEmicro's reset scripts so users will not face this problem.

Best,

Takao

PEmicro Design Engineer

Re: MMU running problem in codewarrior 10.6 environment

martin_kovar — Fri, 14 Dec 2018 11:24:04 GMT

Hello Takao,

thank you very much for explanation. Now, it makes sense to me.

So, it is necessary to edit TCL script.

Regards,

Martin