From mboxrd@z Thu Jan 1 00:00:00 1970 From: Siarhei Siamashka Date: Thu, 15 May 2014 15:34:49 +0300 Subject: [U-Boot] [PATCH v3 4/5] net/designware: reorder struct dw_eth_dev to pack more efficiently. In-Reply-To: <1400092229.15871.19.camel@hastur.hellion.org.uk> References: <1399584366.19277.271.camel@hastur.hellion.org.uk> <1399584395-7832-4-git-send-email-ijc@hellion.org.uk> <1400092229.15871.19.camel@hastur.hellion.org.uk> Message-ID: <20140515153449.4257e104@i7> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On Wed, 14 May 2014 19:30:29 +0100 Ian Campbell wrote: > On Thu, 2014-05-08 at 22:26 +0100, Ian Campbell wrote: > > The {r,t}xbuffs fields also need to be aligned. Previously this was done > > implicitly because they immediately followed the descriptor tables. Make this > > explicit and also move to the head of the struct. > > Looks like I managed to not actually commit the move of the field to the > head of the struct! v3.1 follows.... > > Ian. > > 8<------------ > > From 2937ba01841887317f6792709ed57cb86b5fc0cd Mon Sep 17 00:00:00 2001 > From: Ian Campbell > Date: Thu, 1 May 2014 19:45:15 +0100 > Subject: [PATCH] net/designware: reorder struct dw_eth_dev to pack more > efficiently. > > The {tx,rx}_mac_descrtable fields are aligned to ARCH_DMA_MINALIGN, which could > be 256 or even larger. That means there is a potentially huge hole in the > struct before those fields, so move them to the front where they are better > packed. > > Moving them to the front also helps ensure that so long as dw_eth_dev is > properly aligned (which it is since "net/designware: ensure device private data > is DMA aligned.") the {tx,rx}_mac_descrtable will be too, or at least avoids > having to worry too much about compiler specifics. > > The {r,t}xbuffs fields also need to be aligned. Previously this was done > implicitly because they immediately followed the descriptor tables. Make this > explicit and also move to the head of the struct. > > Signed-off-by: Ian Campbell > Cc: Alexey Brodkin > --- > v3.1: Actually move to the head of the struct, like the commit log said... > v3: Also align tx and rx bufs > --- > drivers/net/designware.h | 11 +++++------ > 1 file changed, 5 insertions(+), 6 deletions(-) > > diff --git a/drivers/net/designware.h b/drivers/net/designware.h > index 382b0c7..de2fdcb 100644 > --- a/drivers/net/designware.h > +++ b/drivers/net/designware.h > @@ -215,15 +215,14 @@ struct dmamacdescr { > #endif > > struct dw_eth_dev { > - u32 interface; > - u32 tx_currdescnum; > - u32 rx_currdescnum; > - > struct dmamacdescr tx_mac_descrtable[CONFIG_TX_DESCR_NUM]; > struct dmamacdescr rx_mac_descrtable[CONFIG_RX_DESCR_NUM]; > + char txbuffs[TX_TOTAL_BUFSIZE] __aligned(ARCH_DMA_MINALIGN); > + char rxbuffs[RX_TOTAL_BUFSIZE] __aligned(ARCH_DMA_MINALIGN); > > - char txbuffs[TX_TOTAL_BUFSIZE]; > - char rxbuffs[RX_TOTAL_BUFSIZE]; > + u32 interface; > + u32 tx_currdescnum; > + u32 rx_currdescnum; > > struct eth_mac_regs *mac_regs_p; > struct eth_dma_regs *dma_regs_p; Thanks for your hard work addressing this nasty tftp breakage regression on the Cubietruck (and on the other ARM hardware, which happens to use the same designware driver). Now everything looks perfect. The whole v3 patch set with this v3.1 reorder fix is Tested-by: Siarhei Siamashka and also if anybody cares Reviewed-by: Siarhei Siamashka -- Best regards, Siarhei Siamashka