2007-10-03 02:51:19

by Arthur Kepner

[permalink] [raw]
Subject: [PATCH 5/5] mthca: allow setting "dmabarrier" on user-allocated memory


Allow setting a "dmabarrier" when the mthca driver registers user-
allocated memory.

Signed-off-by: Arthur Kepner <[email protected]>

---

mthca_provider.c | 7 ++++++-
mthca_user.h | 10 +++++++++-
2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index 17486a4..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, 0);
+ 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;


2007-10-03 02:56:55

by Heikki Orsila

[permalink] [raw]
Subject: Re: [PATCH 5/5] mthca: allow setting "dmabarrier" on user-allocated memory

On Tue, Oct 02, 2007 at 07:50:07PM -0700, [email protected] wrote:
> +struct mthca_reg_mr {
> + __u32 mr_attrs;
> +#define MTHCA_MR_DMAFLUSH 0x1 /* flush in-flight DMA on a write to
> + * memory region */
> + __u32 reserved;
> +};

Seems like a very odd place to #define something new..

--
Heikki Orsila Barbie's law:
[email protected] "Math is hard, let's go shopping!"
http://www.iki.fi/shd

2007-10-03 17:30:36

by Arthur Kepner

[permalink] [raw]
Subject: Re: [PATCH 5/5] mthca: allow setting "dmabarrier" on user-allocated memory

On Wed, Oct 03, 2007 at 05:56:45AM +0300, Heikki Orsila wrote:
> On Tue, Oct 02, 2007 at 07:50:07PM -0700, [email protected] wrote:
> > +struct mthca_reg_mr {
> > + __u32 mr_attrs;
> > +#define MTHCA_MR_DMAFLUSH 0x1 /* flush in-flight DMA on a write to
> > + * memory region */
> > + __u32 reserved;
> > +};
>
> Seems like a very odd place to #define something new..
>

Lots of drivers do that - #define flags or whatever near the
variable to which they apply. Didn't see any examples in the
mthca driver though, so maybe it's setting a precedent there.
If the maintainer objects, I'll move it.

--
Arthur