>-----Original Message----- >From: Ming Lei [mailto:tom.leiming@xxxxxxxxx] >Sent: Thursday, November 25, 2010 6:25 PM >To: Anand Gadiyar >Cc: balbi@xxxxxx; Bob Liu; Zhang, Sonic; Mike Frysinger; >uclinux-dist-devel@xxxxxxxxxxxxxxxxxxxx; Ajay Kumar Gupta; >linux-usb@xxxxxxxxxxxxxxx >Subject: Re: [PATCH] musb_host: Fix high bandwidth iso transfer issue > >2010/11/25 Anand Gadiyar <gadiyar@xxxxxx>: >> Felipe Balbi wrote: >>> Hi, >>> >>> On Thu, Nov 25, 2010 at 04:54:05PM +0800, Bob Liu wrote: >>> >>> The only work around is to config the receive FIFO into single >> buffer >>> >>> mode. >>> >> >>> >> And how to do that ? >>> >> >>> > >>> >It's used to by "USB: musb: disable double buffering for older RTL >>> >versions" >>> >>>http://git.kernel.org/?p=linux/kernel/git/next/linux-next.git >;a=commit >>>dif >> f;h=9f445cb29918dc488b7a9a92ef018599cce33df7 >>> >>> yeah, but that breaks other stuff high bandwidth eps, so we need a >>> better way to "disable double buffering". >>> >> >> Here's how I understand the problem. >> >> Blackfin has a statically configured fifo allocation - with no >> possibility to change it. The IP uses double buffering by >default if a >> FIFO is allocated space enough to hold 2*maxp (which is the case for >> at least some of the endpoints). >> >> The only way to make the IP use single buffering is to tell the IP >> that the maxp is different. >> >> This is an anomaly only in the blackfin - I don't see any >way to work >> around this (for lack of information), than what has been >implemented >> in the patch that already got merged. > >Yes, I agree. > >> >> However, that code should be made to execute only on >blackfin and not >> on other platforms - so that the others are not broken by >that change. > >Even in blackfin platform, the code may break some function, >especially for full speed device, which was explained a lot in >the previous discussion: > > http://marc.info/?t=128576512000002&r=1&w=2 > Hi Ming Lei, This is not a problem for blackfin, because there is no real use case in USB gadget full speed mode from blackfin customers we perceived. But, corrupted RX data in double buffer mode is a critical issue. So, this is an acceptable trade-off for blackfin. Sonic >> >> Is this a fair summary? > > > >-- >Lei Ming > -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
