On Tue, Jan 10, 2017 at 04:56:41PM -0800, Bart Van Assche wrote:
> Several RDMA drivers, e.g. drivers/infiniband/hw/qib, use the CPU to
> transfer data between memory and PCIe adapter. Because of performance
> reasons it is important that the CPU cache is not flushed when such
> drivers transfer data. Make this possible by allowing these drivers to
> override the dma_map_ops pointer. Additionally, introduce the function
> set_dma_ops() that will be used by a later patch in this series.
>
> Signed-off-by: Bart Van Assche <bart.vanassche@xxxxxxxxxxx>
> Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> Cc: Aurelien Jacquiot <a-jacquiot@xxxxxx>
> Cc: Catalin Marinas <catalin.marinas@xxxxxxx>
> Cc: Chris Zankel <chris@xxxxxxxxxx>
> Cc: David Howells <dhowells@xxxxxxxxxx>
> Cc: David S. Miller <davem@xxxxxxxxxxxxx>
> Cc: Fenghua Yu <fenghua.yu@xxxxxxxxx>
> Cc: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> Cc: Geoff Levand <geoff@xxxxxxxxxxxxx>
> Cc: H. Peter Anvin <hpa@xxxxxxxxx>
> Cc: Haavard Skinnemoen <hskinnemoen@xxxxxxxxx>
> Cc: Hans-Christian Egtvedt <egtvedt@xxxxxxxxxxxx>
> Cc: Helge Deller <deller@xxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: James E.J. Bottomley <jejb@xxxxxxxxxxxxxxxx>
> Cc: Jesper Nilsson <jesper.nilsson@xxxxxxxx>
> Cc: Joerg Roedel <joro@xxxxxxxxxx>
> Cc: Jon Mason <jdmason@xxxxxxxx>
> Cc: Jonas Bonn <jonas@xxxxxxxxxxxx>
> Cc: Ley Foon Tan <lftan@xxxxxxxxxx>
> Cc: Mark Salter <msalter@xxxxxxxxxx>
> Cc: Max Filippov <jcmvbkbc@xxxxxxxxx>
> Cc: Mikael Starvik <starvik@xxxxxxxx>
> Cc: Muli Ben-Yehuda <mulix@xxxxxxxxx>
> Cc: Rich Felker <dalias@xxxxxxxx>
> Cc: Russell King <linux@xxxxxxxxxxxxxxx>
> Cc: Stafford Horne <shorne@xxxxxxxxx>
> Cc: Stefan Kristiansson <stefan.kristiansson@xxxxxxxxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Tony Luck <tony.luck@xxxxxxxxx>
> Cc: Will Deacon <will.deacon@xxxxxxx>
> Cc: x86@xxxxxxxxxx
> Cc: Yoshinori Sato <ysato@xxxxxxxxxxxxxxxxxxxx>
> Cc: adi-buildroot-devel@xxxxxxxxxxxxxxxxxxxxx
> Cc: iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx
> Cc: linux-alpha@xxxxxxxxxxxxxxx
> Cc: linux-am33-list@xxxxxxxxxx
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> Cc: linux-c6x-dev@xxxxxxxxxxxxx
> Cc: linux-cris-kernel@xxxxxxxx
> Cc: linux-hexagon@xxxxxxxxxxxxxxx
> Cc: linux-ia64@xxxxxxxxxxxxxxx
> Cc: linux-m68k@xxxxxxxxxxxxxxxxxxxx
> Cc: linux-metag@xxxxxxxxxxxxxxx
> Cc: linux-mips@xxxxxxxxxxxxxx
> Cc: linux-parisc@xxxxxxxxxxxxxxx
> Cc: linux-pci@xxxxxxxxxxxxxxx
> Cc: linux-rdma@xxxxxxxxxxxxxxx
> Cc: linux-s390@xxxxxxxxxxxxxxx
> Cc: linux-sh@xxxxxxxxxxxxxxx
> Cc: linux-snps-arc@xxxxxxxxxxxxxxxxxxx
> Cc: linux-xtensa@xxxxxxxxxxxxxxxx
> Cc: linuxppc-dev@xxxxxxxxxxxxxxxx
> Cc: nios2-dev@xxxxxxxxxxxxxxxxxxxxxx
> Cc: openrisc@xxxxxxxxxxxxxxxxxxxx
> Cc: sparclinux@xxxxxxxxxxxxxxx
> Cc: uclinux-h8-devel@xxxxxxxxxxxxxxxxxxxx
That's a crazy cc: list, you should break this up into smaller pieces,
otherwise it's going to bounce...
> diff --git a/include/linux/device.h b/include/linux/device.h
> index 491b4c0ca633..c7cb225d36b0 100644
> --- a/include/linux/device.h
> +++ b/include/linux/device.h
> @@ -885,6 +885,8 @@ struct dev_links_info {
> * a higher-level representation of the device.
> */
> struct device {
> + const struct dma_map_ops *dma_ops; /* See also get_dma_ops() */
> +
> struct device *parent;
>
> struct device_private *p;
Why not put this new pointer down with the other dma fields in this
structure? Any specific reason it needs to be first?
thanks,
greg k-h
