Re: [U-Boot] [PATCH v5 02/14] net: ftgmac100: add MMU/D-cache support

2013-06-23 Thread Kuo-Jung Su
2013/6/23 Albert ARIBAUD : > Hi Kuo-Jung, > > On Mon, 17 Jun 2013 20:06:52 +0800, Kuo-Jung Su > wrote: > >> 3. No matter if MMU/D-cache is on or off, this patch >>always depends on previous patch: >> >>arm: dma_alloc_coherent: malloc() -> memalign(). >> >>Because the FTGMAC100 expects

Re: [U-Boot] [PATCH v5 02/14] net: ftgmac100: add MMU/D-cache support

2013-06-23 Thread Albert ARIBAUD
Hi Kuo-Jung, On Mon, 17 Jun 2013 20:06:52 +0800, Kuo-Jung Su wrote: > 3. No matter if MMU/D-cache is on or off, this patch >always depends on previous patch: > >arm: dma_alloc_coherent: malloc() -> memalign(). > >Because the FTGMAC100 expects the tx/rx descriptors >are always b

[U-Boot] [PATCH v5 02/14] net: ftgmac100: add MMU/D-cache support

2013-06-17 Thread Kuo-Jung Su
From: Kuo-Jung Su 1. Dynamically allocate the tx/rx descriptors through dma_alloc_coherent(), instead of allocating at compile/link time. 2. Do not block on ftgmac100_send(), the packet has been copied to a safe buffer. Although it's a little bit slower, but the tx performance is