On Tue, 2019-02-05 at 16:36 +-0000, Jason Gunthorpe wrote:
+AD4 On Tue, Feb 05, 2019 at 08:33:44AM -0800, Bart Van Assche wrote:
+AD4 +AD4 On Mon, 2019-02-04 at 19:50 +-0200, Max Gurtovoy wrote:
+AD4 +AD4 +AD4 +AFs ... +AF0
+AD4 +AD4 +AD4 +-/+ACo
+AD4 +AD4 +AD4 +- +ACo struct ib+AF8-scatterlist - Mapped scatterlist for RDMA operations
+AD4 +AD4 +AD4 +- +ACo +AEA-sg: dma mapped sg list
+AD4 +AD4 +AD4 +- +ACo +AEA-dma+AF8-nents: returned by dma+AF8-map+AF8-sg
+AD4 +AD4 +AD4 +- +ACo +AEA-offset: start offset in bytes into the first sg element
+AD4 +AD4 +AD4 +- +ACo-/
+AD4 +AD4 +AD4 +-struct ib+AF8-scatterlist +AHs
+AD4 +AD4 +AD4 +- struct scatterlist +ACo-sg+ADs
+AD4 +AD4 +AD4 +- int dma+AF8-nents+ADs
+AD4 +AD4 +AD4 +- unsigned int offset+ADs
+AD4 +AD4 +AD4 +-+AH0AOw
+AD4 +AD4
+AD4 +AD4 There is nothing in this data structure that is RDMA specific. Are you sure
+AD4 +AD4 the name +ACI-ib+AF8-scatterlist+ACI is a good choice?
+AD4
+AD4 Why do we need an offset anyhow? What is that?
+AD4
+AD4 Is that actually the MR virtual base address?
Hi Jason,
You may want to have a look at the following commit:
commit 9aa8b3217ed3c13d4e3496020b140da0e6f49a08
Author: Bart Van Assche +ADw-bart.vanassche+AEA-sandisk.com+AD4
Date: Thu May 12 10:49:15 2016 -0700
IB/core: Enhance ib+AF8-map+AF8-mr+AF8-sg()
The SRP initiator allows to set max+AF8-sectors to a value that exceeds
the largest amount of data that can be mapped at once with an mlx4
HCA using fast registration and a page size of 4 KB. Hence modify
ib+AF8-map+AF8-mr+AF8-sg() such that it can map partial sg-elements. If an
sg-element has been mapped partially, let the caller know
which fraction has been mapped by adjusting +ACo-sg+AF8-offset.
+AFs .... +AF0
+- +ACo +AEA-sg+AF8-offset+AF8-p: IN: start offset in bytes into sg
+- +ACo OUT: offset in bytes for element n of the sg of the first
+- +ACo byte that has not been processed where n is the return
+- +ACo value of this function.
+AFs .... +AF0
int ib+AF8-sg+AF8-to+AF8-pages(struct ib+AF8-mr +ACo-mr, struct scatterlist +ACo-sgl, int sg+AF8-nents,
- unsigned int sg+AF8-offset, int (+ACo-set+AF8-page)(struct ib+AF8-mr +ACo, u64))
+- unsigned int
+ACo-sg+AF8-offset+AF8-p, int (+ACo-set+AF8-page)(struct ib+AF8-mr +ACo, u64))
+AFs ... +AF0
Bart.
