Anand >From: Anand Gadiyar <vikram.pandita@xxxxxx> > >MUSB DMA_INTR register may sometimes read zero when infact there > >was a pending interrupt. Workaround this by reading the DMA_COUNT >values for all enabled channels when this condition occurs. >Flag these channels as the ones needing to be serviced. > >Signed-off-by: Anand Gadiyar <gadiyar@xxxxxx> >Cc: Ajay Kumar Gupta <ajay.gupta@xxxxxx> >Cc: Felipe Balbi <felipe.balbi@xxxxxxxxx> >Cc: David Brownell <dbrownell@xxxxxxxxxxxxxxxxxxxxx> >Cc: Sergei Shtylyov <sshtylyov@xxxxxxxxxxxxx> >--- >This issue exists on versions of the controller newer than 1.8. >The first version of this patch checked for this version >before acting. However to keep the code clean, I felt that >it is acceptable to do this unconditionally on all versions. > >Also, the absence of the debug print meant that we would >never catch a spurious DMA interrupt on the older versions, >so adding the prints helps there. > > drivers/usb/musb/musbhsdma.c | 24 ++++++++++++++++++++++-- > 1 files changed, 22 insertions(+), 2 deletions(-) > >Index: linux-omap-2.6/drivers/usb/musb/musbhsdma.c >=================================================================== >--- linux-omap-2.6.orig/drivers/usb/musb/musbhsdma.c >+++ linux-omap-2.6/drivers/usb/musb/musbhsdma.c >@@ -256,14 +256,34 @@ static irqreturn_t dma_controller_irq(in > spin_lock_irqsave(&musb->lock, flags); > > int_hsdma = musb_readb(mbase, MUSB_HSDMA_INTR); >- if (!int_hsdma) >- goto done; > > #ifdef CONFIG_BLACKFIN > /* Clear DMA interrupt flags */ > musb_writeb(mbase, MUSB_HSDMA_INTR, int_hsdma); > #endif > >+ if (!int_hsdma) { >+ DBG(2, "spurious DMA irq\n"); >+ >+ for (bchannel = 0; bchannel < MUSB_HSDMA_CHANNELS; bchannel++) { >+ musb_channel = (struct musb_dma_channel *) >+ &(controller->channel[bchannel]); >+ channel = &musb_channel->channel; >+ if (channel->status == MUSB_DMA_STATUS_BUSY) { >+ csr = musb_readw(mbase, >+ MUSB_HSDMA_CHANNEL_OFFSET(bchannel, >+ MUSB_HSDMA_COUNT)); csr is a misnomer. It should be called count. Also I recommend to add a function: musb_read_hsdma_count() in "musbhsdma.h" just like the write counterpart and use here. >+ if (csr == 0) Should this not be if (csr != 0) only then you mark that dma channel as generated the interrupt? >+ int_hsdma |= (1 << bchannel); >+ } >+ } >+ >+ DBG(2, "int_hsdma = 0x%x\n", int_hsdma); >+ >+ if (!int_hsdma) >+ goto done; >+ } >+ > for (bchannel = 0; bchannel < MUSB_HSDMA_CHANNELS; bchannel++) { > if (int_hsdma & (1 << bchannel)) { > musb_channel = (struct musb_dma_channel *) >-- >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 -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html