Re: [PATCH][3/4] usb: ohci-sm501 driver
"Magnus Damm" <[email protected]> Tue, 15 Jan 2008 18:32:08 +0900
| Newsgroups | gmane.linux.usb.devel |
|---|---|
| Message-ID | <[email protected]> |
On Jan 15, 2008 4:54 PM, David Brownell <[email protected]> wrote: > On Sunday 13 January 2008, Magnus Damm wrote: > > usb: ohci-sm501 driver > > > > This patch adds sm501 ohci support. > > Just sm501, not the sm502 you had mentioned? The driver works with sm502 as well. sm501 and sm502 seem very similar. The sm502 device is using the same pci device id as sm501. The sm501 mfd driver works well with sm502 devices - at least for my little setup. > > It's all very straightforward with the > > exception of dma_declare_coherent_memory() and HCD_BOUNCE. > > If I skimmed those docs right, sm502 won't always need that. > > For example, if the sm502 is hooked up through PCI, it doesn't > necessarily even need to have local memory ... or use it for > OHCI, if it's present. (Certainly there's a DMA controller > on the chip that can do PCI bus mastering. I'm assuming that > the OHCI silicon is hooked up to that controller...) You may be right. I'm not sure if it can do PCI bus mastering and if the ohci block is hooked up to the dma controller. I have not played around so much with the PCI interface. My main goal is to add support for usb to some embedded boards made by Renesas. They are hooked up with "slave mode" so they need the bounce buffers. And the bounce buffers _do_ work for pci as well - I just double checked using my sm502 pci card. > If that's so, that stuff should be a function of how the > platform device was configured. For example, an sm502 on > a PCI card might set up a bunch of platform devices for > its various functions (assuming they don't show up as normal > functions) ... but without memory resource #1, thereby > signaling that it doesn't need to use local memory or > do the dma_alloc_coherent() dance. I totally agree. I can probably allocate find time to hack up pci bus mastering support - if it works properly that is. Given that, would it be possible to accept this implementation as a first step? It does work for both "slave mode" and pci, even if pci may be suboptimal. I prefer to see pci bus mastering as an enhancement. / magnus ------------------------------------------------------------------------- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace _______________________________________________ [email protected] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel