Hi Robert, AFAIK, the pxa3xx UDC is merely a bug-fixed version of pxa27x UDC, including DMA request to channel map registers (DRCMRxx). So go on with your excellent work to add DMA support. PS: the DRCMRxx register definitions are now moved into dma.h, and simplified to DRCMR(xx) macros _only_, previous DRCMRxx or DRCMRXXXX are no longer valid. Thanks - eric > -----Original Message----- > From: Robert Jarzmik [mailto:robert.jarzmik@xxxxxxx] > Sent: 2009年3月9日 1:20 > To: Eric Miao; aric@xxxxxxxxxxxxxx > Cc: linux-usb@xxxxxxxxxxxxxxx; David Brownell > Subject: pxa27x_udc and pxa3xx compatibility > > > Hi Eric, > > Aric came up with a little patch (joined at bottom) to make pxa27x_udc work > on a > pxa320 platform. > > As you may know, I have no access to the pxa320 developer's manual. Do you > have > a way to ask at Marvell if both pxa27x and pxa3xx share the same interface > towards the USB Client (UDC), and at least that the pxa27x specification is > valid for a pxa3xx ? And as I'm planning a little extension towards DMA > support, > is there something I should know before I begin so that pxa3xx support comes > "naturally" ? > > From experimentation, Aric reports that pxa27x_udc does work, even the parts > (like RNDIS) which didn't work previously due to buggy pxa27x silicon. > > Cheers. > > -- > Robert > > PS: Aric, would you retest that patch, as I transformed "a bit" your > original > work, and confirm your "Signed-off-by" stills holds for that patch please ? > > From c225a9dcca4523e9d769bd8a6b1e110b57c77521 Mon Sep 17 00:00:00 2001 > From: Aric Blumer <aric@xxxxxxxxxxxxxx> > Date: Sat, 7 Mar 2009 18:36:07 +0100 > Subject: [PATCH] pxa27x_udc: compatibility with pxa320 SoC > > Got pxa27x_udc working on the pxa320 Nomad platform. The > problem was that the pxa3xx UDC is not quite compatible with > the pxa27x UDC in how it handles back-to-back control > packets. The pxa27x probably drops them by default, but the > pxa320 does not, and you have to detect it and set the OPC > bit to clear the zero-length packet. > > Signed-off-by: Aric Blumer <aric@xxxxxxxxxxxxxx> > Signed-off-by: Robert Jarzmik <robert.jarzmik@xxxxxxx> > --- > drivers/usb/gadget/Kconfig | 2 +- > drivers/usb/gadget/pxa27x_udc.c | 20 +++++++++++++++++++- > 2 files changed, 20 insertions(+), 2 deletions(-) > > diff --git a/drivers/usb/gadget/Kconfig b/drivers/usb/gadget/Kconfig > index e55fef5..4e298b3 100644 > --- a/drivers/usb/gadget/Kconfig > +++ b/drivers/usb/gadget/Kconfig > @@ -253,7 +253,7 @@ config USB_PXA25X_SMALL > > config USB_GADGET_PXA27X > boolean "PXA 27x" > - depends on ARCH_PXA && PXA27x > + depends on ARCH_PXA && (PXA27x || PXA3xx) > help > Intel's PXA 27x series XScale ARM v5TE processors include > an integrated full speed USB 1.1 device controller. > diff --git a/drivers/usb/gadget/pxa27x_udc.c > b/drivers/usb/gadget/pxa27x_udc.c > index 990f40f..121dae6 100644 > --- a/drivers/usb/gadget/pxa27x_udc.c > +++ b/drivers/usb/gadget/pxa27x_udc.c > @@ -37,7 +37,10 @@ > #include <linux/usb.h> > #include <linux/usb/ch9.h> > #include <linux/usb/gadget.h> > + > +#ifdef CONFIG_PXA2xx > #include <mach/pxa2xx-regs.h> /* FIXME: for PSSR */ > +#endif > #include <mach/udc.h> > > #include "pxa27x_udc.h" > @@ -1729,6 +1732,17 @@ static void handle_ep0_ctrl_req(struct pxa_udc *udc, > > nuke(ep, -EPROTO); > > + /* > + * In the PXA320 manual, in the section about Back-to-Back setup > + * packets, it describes this situation. The solution is to set OPC > to > + * get rid of the status packet, and then continue with the setup > + * packet. > + */ > + if (cpu_is_pxa3xx()) { > + if (!ep_is_empty(ep) && (ep_count_bytes_remain(ep) == 0)) > + udc_ep_writel(ep, UDCCSR, UDCCSR0_OPC); > + } > + > /* read SETUP packet */ > for (i = 0; i < 2; i++) { > if (unlikely(ep_is_empty(ep))) > @@ -1802,6 +1816,8 @@ stall: > * cleared by software. > * - clearing UDCCSR0_OPC always flushes ep0. If in setup stage, never do > it > * before reading ep0. > + * This is true only for PXA27x. This is not true anymore for PXA3xx > family > + * (check Back-to-Back setup packet in developers guide). > * - irq can be called on a "packet complete" event (opc_irq=1), while > * UDCCSR0_OPC is not yet raised (delta can be as big as 100ms > * from experimentation). > @@ -2356,8 +2372,10 @@ static int pxa_udc_resume(struct platform_device > *_dev) > * Software must configure the USB OTG pad, UDC, and UHC > * to the state they were in before entering sleep mode. > */ > +#ifdef CONFIG_PXA2xx > if (cpu_is_pxa27x()) > PSSR |= PSSR_OTGPH; > +#endif > > return 0; > } > @@ -2381,7 +2399,7 @@ static struct platform_driver udc_driver = { > > static int __init udc_init(void) > { > - if (!cpu_is_pxa27x()) > + if (!cpu_is_pxa27x() && !cpu_is_pxa3xx()) > return -ENODEV; > > printk(KERN_INFO "%s: version %s\n", driver_name, DRIVER_VERSION); > -- > 1.5.6.5 ?韬{.n?????%??檩??w?{.n???{炳???骅w* jg????????G??⒏ ⒎?:+v???????? ????＂??????
