Re: [PATCH] usb: xhci-mtk: set the dma max_seg_size

[Robin Murphy]

On 2023-06-29 09:40, Greg Kroah-Hartman wrote:
> On Thu, Jun 29, 2023 at 09:13:23AM +0200, Ricardo Ribalda wrote:
> > Hi Zubin
> >
> > On Wed, 28 Jun 2023 at 23:57, Zubin Mithra <zsm@xxxxxxxxxxxx> wrote:
> > > On Wed, Jun 28, 2023 at 11:04:20PM +0200, Ricardo Ribalda wrote:
> > > > On Wed, 28 Jun 2023 at 23:00, Ricardo Ribalda <ribalda@xxxxxxxxxxxx> wrote:
> > > > > Allow devices to have dma operations beyond 64K, and avoid warnings such
> > > > > as:
> > > > >
> > > > > DMA-API: xhci-mtk 11200000.usb: mapping sg segment longer than device claims to support [len=98304] [max=65536]
> > > > >
> > > > > Signed-off-by: Ricardo Ribalda <ribalda@xxxxxxxxxxxx>
> > > > Reported-by: Zubin Mithra <zsm@xxxxxxxxxxxx>
> > >
> > > Should this be cc'd to stable@ as well?
> >
> > Not sure, in most of the cases this is "just" a warning fix. Let the
> > maintainer decide:
>
> Warnings can cause reboots as the majority of the linux systems in the
> world run panic-on-warn, so yes, it should be backported.

Although in this particular case, running DMA_API_DEBUG=y on production 
systems is a pretty inadvisable thing to do anyway ;)

However I'm glad I looked, since I think this also points to a bug in 
dma_alloc_noncontiguous() - it's one thing to blame a driver for trying 
to map a malformed scatterlist of its own, but if the DMA API is 
generating one internally without respecting the device's (claimed) 
parameters, then that's on us. I'll have a look into it...

Thanks,
Robin.