Re: [lwip-users] Memory management for packet buffers

[address@hidden]

Mason wrote:
> I'm not sure how to do this within the pbuf infrastructure?
Normally, for a DMA-enabled MAC, you would just pre-allocate 120 
PBUF_POOL pbufs (each 1536 bytes big) and pass the payload pointer to 
the DMA engine. By substracting the difference between 'struct pbuf' and 
its member 'payload', you can then later get the pbuf pointer from the 
buffer pointer. (This is for RX.)

You don't need custom pbufs for that unless you need the memory to be 
located at specific addresses. You can even make the PBUF_POOL pbufs 
bigger than needed and manually adjust its length after allocation so 
that you can put private data into the allocated memory that is not used 
by the pbuf's payload (although that rather counts as a hack, then).
> (I can't mask interrupts just because the network stacks needs a block 
> of memory, or because multiple threads are using the socket layer.)
That depends. I bet the maximum interrupt-disable-time (i.e. interrupt 
latency) won't be lengthened by lwIP (SYS_ARCH_PROTECT is only used 
around small code bits), only frequency of interrupt-disable will increase.

After all, you don't need to globally disable all interrupts, disabling 
the interrupts that use lwIP is enough (i.e. only disable the MAC 
interrupt).

Simon