Use the dma_flags_set_dmabarrier() interface to allow a "dmabarrier"
attribute to be associated with user-allocated memory. (For now,
it's only implemented for mthca.)
Signed-off-by: Arthur Kepner <[email protected]>
---
drivers/infiniband/core/umem.c | 7 +++++--
drivers/infiniband/hw/amso1100/c2_provider.c | 2 +-
drivers/infiniband/hw/cxgb3/iwch_provider.c | 2 +-
drivers/infiniband/hw/ehca/ehca_mrmw.c | 2 +-
drivers/infiniband/hw/ipath/ipath_mr.c | 2 +-
drivers/infiniband/hw/mlx4/cq.c | 2 +-
drivers/infiniband/hw/mlx4/doorbell.c | 2 +-
drivers/infiniband/hw/mlx4/mr.c | 3 ++-
drivers/infiniband/hw/mlx4/qp.c | 2 +-
drivers/infiniband/hw/mlx4/srq.c | 2 +-
drivers/infiniband/hw/mthca/mthca_provider.c | 7 ++++++-
drivers/infiniband/hw/mthca/mthca_user.h | 10 +++++++++-
include/rdma/ib_umem.h | 4 ++--
13 files changed, 32 insertions(+), 15 deletions(-)
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index 664d2fa..5b30b0c 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -69,9 +69,10 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
* @addr: userspace virtual address to start at
* @size: length of region to pin
* @access: IB_ACCESS_xxx flags for memory being pinned
+ * @dmabarrier: set "dmabarrier" attribute on this memory, if necessary
*/
struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
- size_t size, int access)
+ size_t size, int access, int dmabarrier)
{
struct ib_umem *umem;
struct page **page_list;
@@ -83,6 +84,8 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
int ret;
int off;
int i;
+ int flags = dmabarrier ? dma_flags_set_dmabarrier(DMA_BIDIRECTIONAL):
+ DMA_BIDIRECTIONAL;
if (!can_do_mlock())
return ERR_PTR(-EPERM);
@@ -160,7 +163,7 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
chunk->nmap = ib_dma_map_sg(context->device,
&chunk->page_list[0],
chunk->nents,
- DMA_BIDIRECTIONAL);
+ flags);
if (chunk->nmap <= 0) {
for (i = 0; i < chunk->nents; ++i)
put_page(chunk->page_list[i].page);
diff --git a/drivers/infiniband/hw/amso1100/c2_provider.c b/drivers/infiniband/hw/amso1100/c2_provider.c
index 997cf15..17243b7 100644
--- a/drivers/infiniband/hw/amso1100/c2_provider.c
+++ b/drivers/infiniband/hw/amso1100/c2_provider.c
@@ -449,7 +449,7 @@ static struct ib_mr *c2_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
return ERR_PTR(-ENOMEM);
c2mr->pd = c2pd;
- c2mr->umem = ib_umem_get(pd->uobject->context, start, length, acc);
+ c2mr->umem = ib_umem_get(pd->uobject->context, start, length, acc, 0);
if (IS_ERR(c2mr->umem)) {
err = PTR_ERR(c2mr->umem);
kfree(c2mr);
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index f0c7775..d0a514c 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -601,7 +601,7 @@ static struct ib_mr *iwch_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
if (!mhp)
return ERR_PTR(-ENOMEM);
- mhp->umem = ib_umem_get(pd->uobject->context, start, length, acc);
+ mhp->umem = ib_umem_get(pd->uobject->context, start, length, acc, 0);
if (IS_ERR(mhp->umem)) {
err = PTR_ERR(mhp->umem);
kfree(mhp);
diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c b/drivers/infiniband/hw/ehca/ehca_mrmw.c
index d97eda3..c13c11c 100644
--- a/drivers/infiniband/hw/ehca/ehca_mrmw.c
+++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c
@@ -329,7 +329,7 @@ struct ib_mr *ehca_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
}
e_mr->umem = ib_umem_get(pd->uobject->context, start, length,
- mr_access_flags);
+ mr_access_flags, 0);
if (IS_ERR(e_mr->umem)) {
ib_mr = (void *)e_mr->umem;
goto reg_user_mr_exit1;
diff --git a/drivers/infiniband/hw/ipath/ipath_mr.c b/drivers/infiniband/hw/ipath/ipath_mr.c
index e442470..e351222 100644
--- a/drivers/infiniband/hw/ipath/ipath_mr.c
+++ b/drivers/infiniband/hw/ipath/ipath_mr.c
@@ -195,7 +195,7 @@ struct ib_mr *ipath_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
goto bail;
}
- umem = ib_umem_get(pd->uobject->context, start, length, mr_access_flags);
+ umem = ib_umem_get(pd->uobject->context, start, length, mr_access_flags, 0);
if (IS_ERR(umem))
return (void *) umem;
diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c
index 8bf44da..96ec149 100644
--- a/drivers/infiniband/hw/mlx4/cq.c
+++ b/drivers/infiniband/hw/mlx4/cq.c
@@ -122,7 +122,7 @@ struct ib_cq *mlx4_ib_create_cq(struct ib_device *ibdev, int entries, int vector
}
cq->umem = ib_umem_get(context, ucmd.buf_addr, buf_size,
- IB_ACCESS_LOCAL_WRITE);
+ IB_ACCESS_LOCAL_WRITE, 0);
if (IS_ERR(cq->umem)) {
err = PTR_ERR(cq->umem);
goto err_cq;
diff --git a/drivers/infiniband/hw/mlx4/doorbell.c b/drivers/infiniband/hw/mlx4/doorbell.c
index 1c36087..0afde2d 100644
--- a/drivers/infiniband/hw/mlx4/doorbell.c
+++ b/drivers/infiniband/hw/mlx4/doorbell.c
@@ -181,7 +181,7 @@ int mlx4_ib_db_map_user(struct mlx4_ib_ucontext *context, unsigned long virt,
page->user_virt = (virt & PAGE_MASK);
page->refcnt = 0;
page->umem = ib_umem_get(&context->ibucontext, virt & PAGE_MASK,
- PAGE_SIZE, 0);
+ PAGE_SIZE, 0, 0);
if (IS_ERR(page->umem)) {
err = PTR_ERR(page->umem);
kfree(page);
diff --git a/drivers/infiniband/hw/mlx4/mr.c b/drivers/infiniband/hw/mlx4/mr.c
index 85ae906..3e05eb6 100644
--- a/drivers/infiniband/hw/mlx4/mr.c
+++ b/drivers/infiniband/hw/mlx4/mr.c
@@ -133,7 +133,8 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
if (!mr)
return ERR_PTR(-ENOMEM);
- mr->umem = ib_umem_get(pd->uobject->context, start, length, access_flags);
+ mr->umem = ib_umem_get(pd->uobject->context, start, length,
+ access_flags, 0);
if (IS_ERR(mr->umem)) {
err = PTR_ERR(mr->umem);
goto err_free;
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 85c51bd..f4dd553 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -335,7 +335,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
goto err;
qp->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr,
- qp->buf_size, 0);
+ qp->buf_size, 0, 0);
if (IS_ERR(qp->umem)) {
err = PTR_ERR(qp->umem);
goto err;
diff --git a/drivers/infiniband/hw/mlx4/srq.c b/drivers/infiniband/hw/mlx4/srq.c
index 408748f..7c17b89 100644
--- a/drivers/infiniband/hw/mlx4/srq.c
+++ b/drivers/infiniband/hw/mlx4/srq.c
@@ -115,7 +115,7 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
}
srq->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr,
- buf_size, 0);
+ buf_size, 0, 0);
if (IS_ERR(srq->umem)) {
err = PTR_ERR(srq->umem);
goto err_srq;
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index 6bcde1c..c818708 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -1017,17 +1017,22 @@ static struct ib_mr *mthca_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
struct mthca_dev *dev = to_mdev(pd->device);
struct ib_umem_chunk *chunk;
struct mthca_mr *mr;
+ struct mthca_reg_mr ucmd;
u64 *pages;
int shift, n, len;
int i, j, k;
int err = 0;
int write_mtt_size;
+ if (ib_copy_from_udata(&ucmd, udata, sizeof ucmd))
+ return ERR_PTR(-EFAULT);
+
mr = kmalloc(sizeof *mr, GFP_KERNEL);
if (!mr)
return ERR_PTR(-ENOMEM);
- mr->umem = ib_umem_get(pd->uobject->context, start, length, acc);
+ mr->umem = ib_umem_get(pd->uobject->context, start, length, acc,
+ ucmd.mr_attrs & MTHCA_MR_DMAFLUSH);
if (IS_ERR(mr->umem)) {
err = PTR_ERR(mr->umem);
goto err;
diff --git a/drivers/infiniband/hw/mthca/mthca_user.h b/drivers/infiniband/hw/mthca/mthca_user.h
index 02cc0a7..5662aea 100644
--- a/drivers/infiniband/hw/mthca/mthca_user.h
+++ b/drivers/infiniband/hw/mthca/mthca_user.h
@@ -41,7 +41,7 @@
* Increment this value if any changes that break userspace ABI
* compatibility are made.
*/
-#define MTHCA_UVERBS_ABI_VERSION 1
+#define MTHCA_UVERBS_ABI_VERSION 2
/*
* Make sure that all structs defined in this file remain laid out so
@@ -61,6 +61,14 @@ struct mthca_alloc_pd_resp {
__u32 reserved;
};
+struct mthca_reg_mr {
+ __u32 mr_attrs;
+#define MTHCA_MR_DMAFLUSH 0x1 /* flush in-flight DMA on a write to
+ * memory region */
+ __u32 reserved;
+};
+
+
struct mthca_create_cq {
__u32 lkey;
__u32 pdn;
diff --git a/include/rdma/ib_umem.h b/include/rdma/ib_umem.h
index c533d6c..054d279 100644
--- a/include/rdma/ib_umem.h
+++ b/include/rdma/ib_umem.h
@@ -61,7 +61,7 @@ struct ib_umem_chunk {
#ifdef CONFIG_INFINIBAND_USER_MEM
struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
- size_t size, int access);
+ size_t size, int access, int dmabarrier);
void ib_umem_release(struct ib_umem *umem);
int ib_umem_page_count(struct ib_umem *umem);
@@ -71,7 +71,7 @@ int ib_umem_page_count(struct ib_umem *umem);
static inline struct ib_umem *ib_umem_get(struct ib_ucontext *context,
unsigned long addr, size_t size,
- int access) {
+ int access, int dmabarrier) {
return ERR_PTR(-EINVAL);
}
static inline void ib_umem_release(struct ib_umem *umem) { }
Sorry for not mentioning this earlier, but this patch should really be
two (or more) patches: one to add dmabarrier support to the core user
memory stuff in drivers/infiniband, and a second one to add support to
mthca (and more patches to add support to mlx4, cxgb3, etc, etc).
> + * @dmabarrier: set "dmabarrier" attribute on this memory, if necessary
Nit: just delete the "if necessary" since I don't think it makes
things clearer (and actually doesn't make much sense in this context)
Other than that this look fine to me, and I'm ready to merge it once
the necessary core DMA stuff is settled.
- R.
On Fri, Sep 28, 2007 at 12:50:00PM -0700, Roland Dreier wrote:
> Sorry for not mentioning this earlier, but this patch should really be
> two (or more) patches: one to add dmabarrier support to the core user
> memory stuff in drivers/infiniband, and a second one to add support to
> mthca (and more patches to add support to mlx4, cxgb3, etc, etc).
Makes sense.
>
> > + * @dmabarrier: set "dmabarrier" attribute on this memory, if necessary
>
> Nit: just delete the "if necessary" since I don't think it makes
> things clearer (and actually doesn't make much sense in this context)
>
OK.
> Other than that this look fine to me, and I'm ready to merge it once
> the necessary core DMA stuff is settled.
>
Great. A new version of the patchset is on the way.
--
Arthur