[lwip-users] Ang: Re: Bare Metal / Memory Alignment
From:
ake . forslund
Subject:
[lwip-users] Ang: Re: Bare Metal / Memory Alignment
Date:
Wed, 7 Sep 2011 17:59:01 +0200
I recently got the impression that structure-packing wasn't mandatory and could be replaced with padding, I'm not at all sure about this so Stick with packing until Kieran or Simon says otherwise.
I've set up packing a bit differently, firstly I don't use bpstruct.h/epstruct.h at all and my compiler hints for packing looks like this
cc.h:
/* Compiler hints for packing structures */ #define PACK_STRUCT_FIELD(x) x #define PACK_STRUCT_STRUCT #define PACK_STRUCT_BEGIN __packed #define PACK_STRUCT_END
I'm not sure I'm helping at all, but I like to try anyway =)
/Åke
----------------------------- Åke Forslund 0433-273296 NIBE AB Box 14 S-285 21 Markaryd Tel +46-(0)433-273296
address@hidden skrev: -----
Till: Mailing list for lwIP users <address@hidden> Från: Jens Breuer Sänt av: address@hidden Datum: 2011-09-07 16:16 Ärende: Re: [lwip-users] Bare Metal / Memory Alignment
Hi Ake,
> My Target cpu is an Atmel AT91SAM9261 (ARM9), and I haven't seen any problems like the one you described. I'm not sure I had exactly the same problems but when I started to work with lwip it discarded packages due to missing settings for packing structs.
The target device I am using is STM32F107VC.
As explained there are no other errors so far. At least none that I am aware of at this time. However I am still quite at the beginning. Next I'm going to get TCP and UDP included and test that.
> Are you using packed structs or ETH_PAD_SIZE set to some value? Just guesses from the top of my head.
Yes, I use packed structs. I thought, that it would be mandatory. These are the related defines in cc.h: ===3<=== #define PACK_STRUCT_USE_INCLUDES #define PACK_STRUCT_FIELD(x) x #define PACK_STRUCT_STRUCT #define PACK_STRUCT_BEGIN #define PACK_STRUCT_END ===3<===
This is the content of bpstruct.h: ===3<=== #pragma push #pragma pack(1) ===3<===
This is the content of epstruct.h: ===3<=== #pragma pop ===3<===
ETH_PAD_SIZE is not defined in my case. It is using the default from opt.h of 0. I set it to 2 some time, but I realized quickly that lwIP then uses the wrong data as the type field value (misalign by 2, surprise).
However, I still have the feeling that I am just doing something wrong...
@Kieran & Ake: Thanks for your replies and interest in the problem.