Re: [PATCH 5.1 29/96] usb: gadget: udc: lpc32xx: allocate descriptor with GFP_ATOMIC

[Alexandre Belloni]

On 10/07/2019 11:59:10+0200, Pavel Machek wrote:
> On Mon 2019-07-08 17:13:01, Greg Kroah-Hartman wrote:
> > [ Upstream commit fbc318afadd6e7ae2252d6158cf7d0c5a2132f7d ]
> > 
> > Gadget drivers may queue request in interrupt context. This would lead to
> > a descriptor allocation in that context. In that case we would hit
> > BUG_ON(in_interrupt()) in __get_vm_area_node.
> > 
> > Also remove the unnecessary cast.
> 
> GFP_ATOMIC allocations can and do fail, but I don't see any explicit error handling.
> 
> Can someone check everything is ok?
> 

It is checked later on:

		dd = udc_dd_alloc(udc);
		if (!dd) {
			/* Error allocating DD */
			return -ENOMEM;
		}

> Thanks,
> 									Pavel
> 
> 
> > diff --git a/drivers/usb/gadget/udc/lpc32xx_udc.c b/drivers/usb/gadget/udc/lpc32xx_udc.c
> > index b0781771704e..eafc2a00c96a 100644
> > --- a/drivers/usb/gadget/udc/lpc32xx_udc.c
> > +++ b/drivers/usb/gadget/udc/lpc32xx_udc.c
> > @@ -922,8 +922,7 @@ static struct lpc32xx_usbd_dd_gad *udc_dd_alloc(struct lpc32xx_udc *udc)
> >  	dma_addr_t			dma;
> >  	struct lpc32xx_usbd_dd_gad	*dd;
> >  
> > -	dd = (struct lpc32xx_usbd_dd_gad *) dma_pool_alloc(
> > -			udc->dd_cache, (GFP_KERNEL | GFP_DMA), &dma);
> > +	dd = dma_pool_alloc(udc->dd_cache, GFP_ATOMIC | GFP_DMA, &dma);
> >  	if (dd)
> >  		dd->this_dma = dma;
> >  
> > -- 
> > 2.20.1
> > 
> > 
> 
> -- 
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

-- 
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com