Re: [PATCH net] sfc: Reduce RX scatter buffer size to multiple of256

[Ben Hutchings]

On Mon, 2013-05-13 at 21:02 +0200, Geert Uytterhoeven wrote:
> Hi Ben,
> 
> On Mon, May 13, 2013 at 7:48 PM, Ben Hutchings
> <bhutchings@xxxxxxxxxxxxxx> wrote:
> > efx_start_datapath() asserts that we can fit 2 RX scatter buffers plus
> > a software structure, each cache-aligned, into a single page.  Where
> > L1_CACHE_BYTES == 256 and PAGE_SIZE == 4096, which is the case on
> > s390, this assertion fails.  Reduce EFX_RX_USR_BUF_SIZE to make this
> > work.
> >
> > This should also be good for performance, as it ensures that each RX
> > scatter buffer covers whole cache lines and slightly reduces the use
> > of DMA writes that can require a read-modify-write on inter-processor
> > links.
> >
> > (We could use 2048 - L1_CACHE_BYTES, but EFX_RX_USR_BUF_SIZE also
> > affects user-level networking where a larger amount of housekeeping
> > data may be needed.  Although this version of the driver does not
> > support user-level networking, I prefer to keep scattering behaviour
> > consistent with the out-of-tree version.)
> >
> > Reported-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> > Reported-by: Heiko Carstens <heiko.carstens@xxxxxxxxxx>
> > Signed-off-by: Ben Hutchings <bhutchings@xxxxxxxxxxxxxx>
> > ---
> > Heiko or Geert, please confirm that this really does fix the build
> 
> Thanks! But unfortunately I still get the same error:
> 
>   CC [M]  drivers/net/ethernet/sfc/efx.o
> /scratch/geert/linux/linux-m68k/drivers/net/ethernet/sfc/efx.c: In
> function 'efx_start_datapath':
> /scratch/geert/linux/linux-m68k/drivers/net/ethernet/sfc/efx.c:646:3:
> error: call to '__compiletime_assert_648' declared with attribute
> error: BUILD_BUG_ON failed: sizeof(struct efx_rx_page_state) +
> EFX_PAGE_IP_ALIGN + EFX_RX_USR_BUF_SIZE > PAGE_SIZE / 2
> make[4]: *** [drivers/net/ethernet/sfc/efx.o] Error 1

OK, this doesn't work because on s390 EFX_PAGE_IP_ALIGN == 2 (this macro
is equivalent to NET_IP_ALIGN, though that wasn't always true).  So DMA
is going to be misaligned on s390 anyway.

I'll see if I can work out a sensible definition that works for s390
while still being good for x86 and powerpc (which are the two we mostly
care about).

> > failure - I don't have an s390 toolchain.
> 
> http://kernel.org/pub/tools/crosstool/files/bin/

I know, but this still takes time to set up.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/