* [PATCH for-next 1/9] RDMA/rxe: Add bind MW fields to rxe_send_wr
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-13 23:11 ` Jason Gunthorpe
2021-04-08 21:40 ` [PATCH for-next 2/9] RDMA/rxe: Return errors for add index and key Bob Pearson
` (8 subsequent siblings)
9 siblings, 1 reply; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Add fields to struct rxe_send_wr in rdma_user_rxe.h to
support bind MW work requests from user space and kernel.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
include/uapi/rdma/rdma_user_rxe.h | 34 ++++++++++++++++++++++++++++++-
1 file changed, 33 insertions(+), 1 deletion(-)
diff --git a/include/uapi/rdma/rdma_user_rxe.h b/include/uapi/rdma/rdma_user_rxe.h
index 068433e2229d..b9c80ca73473 100644
--- a/include/uapi/rdma/rdma_user_rxe.h
+++ b/include/uapi/rdma/rdma_user_rxe.h
@@ -99,7 +99,39 @@ struct rxe_send_wr {
__u32 remote_qkey;
__u16 pkey_index;
} ud;
- /* reg is only used by the kernel and is not part of the uapi */
+ struct {
+ __aligned_u64 addr;
+ __aligned_u64 length;
+ union {
+ __u32 mr_lkey;
+ __aligned_u64 reserved1;
+ };
+ union {
+ __u32 mw_rkey;
+ __aligned_u64 reserved2;
+ };
+ __u32 rkey;
+ __u32 access;
+ __u32 flags;
+ } umw;
+ /* The following are only used by the kernel
+ * and are not part of the uapi
+ */
+ struct {
+ __aligned_u64 addr;
+ __aligned_u64 length;
+ union {
+ struct ib_mr *mr;
+ __aligned_u64 reserved1;
+ };
+ union {
+ struct ib_mw *mw;
+ __aligned_u64 reserved2;
+ };
+ __u32 rkey;
+ __u32 access;
+ __u32 flags;
+ } kmw;
struct {
union {
struct ib_mr *mr;
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH for-next 1/9] RDMA/rxe: Add bind MW fields to rxe_send_wr
2021-04-08 21:40 ` [PATCH for-next 1/9] RDMA/rxe: Add bind MW fields to rxe_send_wr Bob Pearson
@ 2021-04-13 23:11 ` Jason Gunthorpe
2021-04-14 2:50 ` Bob Pearson
0 siblings, 1 reply; 13+ messages in thread
From: Jason Gunthorpe @ 2021-04-13 23:11 UTC (permalink / raw)
To: Bob Pearson; +Cc: xyzxyz2000, linux-rdma, Bob Pearson
On Thu, Apr 08, 2021 at 04:40:33PM -0500, Bob Pearson wrote:
> Add fields to struct rxe_send_wr in rdma_user_rxe.h to
> support bind MW work requests from user space and kernel.
>
> Signed-off-by: Bob Pearson <rpearson@hpe.com>
> include/uapi/rdma/rdma_user_rxe.h | 34 ++++++++++++++++++++++++++++++-
> 1 file changed, 33 insertions(+), 1 deletion(-)
>
> diff --git a/include/uapi/rdma/rdma_user_rxe.h b/include/uapi/rdma/rdma_user_rxe.h
> index 068433e2229d..b9c80ca73473 100644
> +++ b/include/uapi/rdma/rdma_user_rxe.h
> @@ -99,7 +99,39 @@ struct rxe_send_wr {
> __u32 remote_qkey;
> __u16 pkey_index;
> } ud;
> - /* reg is only used by the kernel and is not part of the uapi */
> + struct {
> + __aligned_u64 addr;
> + __aligned_u64 length;
> + union {
> + __u32 mr_lkey;
> + __aligned_u64 reserved1;
> + };
> + union {
> + __u32 mw_rkey;
> + __aligned_u64 reserved2;
> + };
> + __u32 rkey;
> + __u32 access;
> + __u32 flags;
> + } umw;
> + /* The following are only used by the kernel
> + * and are not part of the uapi
> + */
> + struct {
> + __aligned_u64 addr;
> + __aligned_u64 length;
> + union {
> + struct ib_mr *mr;
A kernel struct should not appear in a uapi header, nor should secure
kernel data ever be stored in user memory.
I'm completely lost at how all this thinks it is working, we don't
even have kernel verbs support for memory windows??
Jason
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [PATCH for-next 1/9] RDMA/rxe: Add bind MW fields to rxe_send_wr
2021-04-13 23:11 ` Jason Gunthorpe
@ 2021-04-14 2:50 ` Bob Pearson
0 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-14 2:50 UTC (permalink / raw)
To: Jason Gunthorpe; +Cc: xyzxyz2000, linux-rdma, Bob Pearson
On 4/13/21 6:11 PM, Jason Gunthorpe wrote:
> On Thu, Apr 08, 2021 at 04:40:33PM -0500, Bob Pearson wrote:
>> Add fields to struct rxe_send_wr in rdma_user_rxe.h to
>> support bind MW work requests from user space and kernel.
>>
>> Signed-off-by: Bob Pearson <rpearson@hpe.com>
>> include/uapi/rdma/rdma_user_rxe.h | 34 ++++++++++++++++++++++++++++++-
>> 1 file changed, 33 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/uapi/rdma/rdma_user_rxe.h b/include/uapi/rdma/rdma_user_rxe.h
>> index 068433e2229d..b9c80ca73473 100644
>> +++ b/include/uapi/rdma/rdma_user_rxe.h
>> @@ -99,7 +99,39 @@ struct rxe_send_wr {
>> __u32 remote_qkey;
>> __u16 pkey_index;
>> } ud;
>> - /* reg is only used by the kernel and is not part of the uapi */
>> + struct {
>> + __aligned_u64 addr;
>> + __aligned_u64 length;
>> + union {
>> + __u32 mr_lkey;
>> + __aligned_u64 reserved1;
>> + };
>> + union {
>> + __u32 mw_rkey;
>> + __aligned_u64 reserved2;
>> + };
>> + __u32 rkey;
>> + __u32 access;
>> + __u32 flags;
>> + } umw;
>> + /* The following are only used by the kernel
>> + * and are not part of the uapi
>> + */
>> + struct {
>> + __aligned_u64 addr;
>> + __aligned_u64 length;
>> + union {
>> + struct ib_mr *mr;
>
> A kernel struct should not appear in a uapi header, nor should secure
> kernel data ever be stored in user memory.
>
> I'm completely lost at how all this thinks it is working, we don't
> even have kernel verbs support for memory windows??
>
> Jason
>
Hard to see from context but I was copying the existing header file which is used to define the WQE layout
for rxe for *both* user and kernel space. The structs shown in the hunk are contained in a union.
The existing version of the driver has one struct
union {
...
struct {
...
} reg;
} wr;
which is only used in kernel space. Like the old song about the 8x10 glossy pictures, I figured I'd just
add another one and avoid looking up the MWs and MRs for bind in the kernel. This is before I discovered
that you all deleted MW support from the kernel a while back. Not really sure why. Not used I suppose but
it is in the spec.
Clearly only user data for user WRs appear in this for user QPs and only kernel data for kernel QPs.
There isn't any mixing. Since it is a union there is no overhead by including the kernel WQEs. There is no
secret about what the kernel structs are.
How to fix this? Since rxe was written the new variable size WQEs in ib_verbs.h (ib_send_wr and friends) were
added and they could be a better choice. Rxe could take these as its own native wqe and that would reduce
the amount of work on kernel WRs and reduce the amount of code dealing with them from user space. It would
break ABI unfortunately.
The other approach is to, as you suggest, ifdef out the kernel stuff in user space. It works anyway because
you can always add a pointer to an undefined struct as long as you don't try to use it.
Also as you point out ib_bind_mw doesn't exist anymore so I could get rid of the kmw struct and just use the
rkeys. Could still support a kernel mw implementation later without a change if one showed up. There are some
interesting articles about the security of RDMA wire protocols that strongly suggest using type2 MWs with
randomly chosen rkeys and qp numbers instead of MRs. It's not perfect either but orders of magnitude stronger
protection.
bob
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH for-next 2/9] RDMA/rxe: Return errors for add index and key
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 1/9] RDMA/rxe: Add bind MW fields to rxe_send_wr Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 3/9] RDMA/rxe: Enable MW object pool Bob Pearson
` (7 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Modify rxe_add_index() and rxe_add_key() to return an
error if the index or key is aleady present in the pool.
Currently they print a warning and silently fail with
bad consequences to the caller.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe_pool.c | 44 ++++++++++++++++++----------
drivers/infiniband/sw/rxe/rxe_pool.h | 8 ++---
2 files changed, 32 insertions(+), 20 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index d24901f2af3f..2b795e2fc4b3 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -183,7 +183,7 @@ static u32 alloc_index(struct rxe_pool *pool)
return index + pool->index.min_index;
}
-static void insert_index(struct rxe_pool *pool, struct rxe_pool_entry *new)
+static int insert_index(struct rxe_pool *pool, struct rxe_pool_entry *new)
{
struct rb_node **link = &pool->index.tree.rb_node;
struct rb_node *parent = NULL;
@@ -195,7 +195,7 @@ static void insert_index(struct rxe_pool *pool, struct rxe_pool_entry *new)
if (elem->index == new->index) {
pr_warn("element already exists!\n");
- goto out;
+ return -EINVAL;
}
if (elem->index > new->index)
@@ -206,11 +206,11 @@ static void insert_index(struct rxe_pool *pool, struct rxe_pool_entry *new)
rb_link_node(&new->index_node, parent, link);
rb_insert_color(&new->index_node, &pool->index.tree);
-out:
- return;
+
+ return 0;
}
-static void insert_key(struct rxe_pool *pool, struct rxe_pool_entry *new)
+static int insert_key(struct rxe_pool *pool, struct rxe_pool_entry *new)
{
struct rb_node **link = &pool->key.tree.rb_node;
struct rb_node *parent = NULL;
@@ -226,7 +226,7 @@ static void insert_key(struct rxe_pool *pool, struct rxe_pool_entry *new)
if (cmp == 0) {
pr_warn("key already exists!\n");
- goto out;
+ return -EINVAL;
}
if (cmp > 0)
@@ -237,26 +237,32 @@ static void insert_key(struct rxe_pool *pool, struct rxe_pool_entry *new)
rb_link_node(&new->key_node, parent, link);
rb_insert_color(&new->key_node, &pool->key.tree);
-out:
- return;
+
+ return 0;
}
-void __rxe_add_key_locked(struct rxe_pool_entry *elem, void *key)
+int __rxe_add_key_locked(struct rxe_pool_entry *elem, void *key)
{
struct rxe_pool *pool = elem->pool;
+ int err;
memcpy((u8 *)elem + pool->key.key_offset, key, pool->key.key_size);
- insert_key(pool, elem);
+ err = insert_key(pool, elem);
+
+ return err;
}
-void __rxe_add_key(struct rxe_pool_entry *elem, void *key)
+int __rxe_add_key(struct rxe_pool_entry *elem, void *key)
{
struct rxe_pool *pool = elem->pool;
unsigned long flags;
+ int err;
write_lock_irqsave(&pool->pool_lock, flags);
- __rxe_add_key_locked(elem, key);
+ err = __rxe_add_key_locked(elem, key);
write_unlock_irqrestore(&pool->pool_lock, flags);
+
+ return err;
}
void __rxe_drop_key_locked(struct rxe_pool_entry *elem)
@@ -276,22 +282,28 @@ void __rxe_drop_key(struct rxe_pool_entry *elem)
write_unlock_irqrestore(&pool->pool_lock, flags);
}
-void __rxe_add_index_locked(struct rxe_pool_entry *elem)
+int __rxe_add_index_locked(struct rxe_pool_entry *elem)
{
struct rxe_pool *pool = elem->pool;
+ int err;
elem->index = alloc_index(pool);
- insert_index(pool, elem);
+ err = insert_index(pool, elem);
+
+ return err;
}
-void __rxe_add_index(struct rxe_pool_entry *elem)
+int __rxe_add_index(struct rxe_pool_entry *elem)
{
struct rxe_pool *pool = elem->pool;
unsigned long flags;
+ int err;
write_lock_irqsave(&pool->pool_lock, flags);
- __rxe_add_index_locked(elem);
+ err = __rxe_add_index_locked(elem);
write_unlock_irqrestore(&pool->pool_lock, flags);
+
+ return err;
}
void __rxe_drop_index_locked(struct rxe_pool_entry *elem)
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.h b/drivers/infiniband/sw/rxe/rxe_pool.h
index 61210b300a78..1feca1bffced 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.h
+++ b/drivers/infiniband/sw/rxe/rxe_pool.h
@@ -111,11 +111,11 @@ int __rxe_add_to_pool(struct rxe_pool *pool, struct rxe_pool_entry *elem);
/* assign an index to an indexed object and insert object into
* pool's rb tree holding and not holding the pool_lock
*/
-void __rxe_add_index_locked(struct rxe_pool_entry *elem);
+int __rxe_add_index_locked(struct rxe_pool_entry *elem);
#define rxe_add_index_locked(obj) __rxe_add_index_locked(&(obj)->pelem)
-void __rxe_add_index(struct rxe_pool_entry *elem);
+int __rxe_add_index(struct rxe_pool_entry *elem);
#define rxe_add_index(obj) __rxe_add_index(&(obj)->pelem)
@@ -133,11 +133,11 @@ void __rxe_drop_index(struct rxe_pool_entry *elem);
/* assign a key to a keyed object and insert object into
* pool's rb tree holding and not holding pool_lock
*/
-void __rxe_add_key_locked(struct rxe_pool_entry *elem, void *key);
+int __rxe_add_key_locked(struct rxe_pool_entry *elem, void *key);
#define rxe_add_key_locked(obj, key) __rxe_add_key_locked(&(obj)->pelem, key)
-void __rxe_add_key(struct rxe_pool_entry *elem, void *key);
+int __rxe_add_key(struct rxe_pool_entry *elem, void *key);
#define rxe_add_key(obj, key) __rxe_add_key(&(obj)->pelem, key)
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 3/9] RDMA/rxe: Enable MW object pool
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 1/9] RDMA/rxe: Add bind MW fields to rxe_send_wr Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 2/9] RDMA/rxe: Return errors for add index and key Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 4/9] RDMA/rxe: Add ib_alloc_mw and ib_dealloc_mw verbs Bob Pearson
` (6 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Currently the rxe driver has a rxe_mw struct object but
nothing about memory windows is enabled. This patch
turns on memory windows and some minor cleanup.
Set device attribute in rxe.c so max_mw = MAX_MW.
Change parameters in rxe_param.h so that MAX_MW is the same as MAX_MR.
Reduce the number of MRs and MWs to 4K from 256K.
Add device capability bits for 2a and 2b memory windows.
Removed RXE_MR_TYPE_MW from the rxe_mr_type enum.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe.c | 1 +
drivers/infiniband/sw/rxe/rxe_param.h | 19 ++++++++++++-------
drivers/infiniband/sw/rxe/rxe_verbs.h | 1 -
3 files changed, 13 insertions(+), 8 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe.c b/drivers/infiniband/sw/rxe/rxe.c
index 95f0de0c8b49..8e0f9c489cab 100644
--- a/drivers/infiniband/sw/rxe/rxe.c
+++ b/drivers/infiniband/sw/rxe/rxe.c
@@ -54,6 +54,7 @@ static void rxe_init_device_param(struct rxe_dev *rxe)
rxe->attr.max_cq = RXE_MAX_CQ;
rxe->attr.max_cqe = (1 << RXE_MAX_LOG_CQE) - 1;
rxe->attr.max_mr = RXE_MAX_MR;
+ rxe->attr.max_mw = RXE_MAX_MW;
rxe->attr.max_pd = RXE_MAX_PD;
rxe->attr.max_qp_rd_atom = RXE_MAX_QP_RD_ATOM;
rxe->attr.max_res_rd_atom = RXE_MAX_RES_RD_ATOM;
diff --git a/drivers/infiniband/sw/rxe/rxe_param.h b/drivers/infiniband/sw/rxe/rxe_param.h
index 25ab50d9b7c2..742e6ec93686 100644
--- a/drivers/infiniband/sw/rxe/rxe_param.h
+++ b/drivers/infiniband/sw/rxe/rxe_param.h
@@ -37,7 +37,6 @@ static inline enum ib_mtu eth_mtu_int_to_enum(int mtu)
enum rxe_device_param {
RXE_MAX_MR_SIZE = -1ull,
RXE_PAGE_SIZE_CAP = 0xfffff000,
- RXE_MAX_QP = 0x10000,
RXE_MAX_QP_WR = 0x4000,
RXE_DEVICE_CAP_FLAGS = IB_DEVICE_BAD_PKEY_CNTR
| IB_DEVICE_BAD_QKEY_CNTR
@@ -49,7 +48,10 @@ enum rxe_device_param {
| IB_DEVICE_RC_RNR_NAK_GEN
| IB_DEVICE_SRQ_RESIZE
| IB_DEVICE_MEM_MGT_EXTENSIONS
- | IB_DEVICE_ALLOW_USER_UNREG,
+ | IB_DEVICE_ALLOW_USER_UNREG
+ | IB_DEVICE_MEM_WINDOW
+ | IB_DEVICE_MEM_WINDOW_TYPE_2A
+ | IB_DEVICE_MEM_WINDOW_TYPE_2B,
RXE_MAX_SGE = 32,
RXE_MAX_WQE_SIZE = sizeof(struct rxe_send_wqe) +
sizeof(struct ib_sge) * RXE_MAX_SGE,
@@ -58,7 +60,6 @@ enum rxe_device_param {
RXE_MAX_SGE_RD = 32,
RXE_MAX_CQ = 16384,
RXE_MAX_LOG_CQE = 15,
- RXE_MAX_MR = 256 * 1024,
RXE_MAX_PD = 0x7ffc,
RXE_MAX_QP_RD_ATOM = 128,
RXE_MAX_RES_RD_ATOM = 0x3f000,
@@ -67,7 +68,6 @@ enum rxe_device_param {
RXE_MAX_MCAST_QP_ATTACH = 56,
RXE_MAX_TOT_MCAST_QP_ATTACH = 0x70000,
RXE_MAX_AH = 100,
- RXE_MAX_SRQ = 960,
RXE_MAX_SRQ_WR = 0x4000,
RXE_MIN_SRQ_WR = 1,
RXE_MAX_SRQ_SGE = 27,
@@ -80,16 +80,21 @@ enum rxe_device_param {
RXE_NUM_PORT = 1,
+ RXE_MAX_QP = 0x10000,
RXE_MIN_QP_INDEX = 16,
RXE_MAX_QP_INDEX = 0x00020000,
+ RXE_MAX_SRQ = 0x00001000,
RXE_MIN_SRQ_INDEX = 0x00020001,
RXE_MAX_SRQ_INDEX = 0x00040000,
+ RXE_MAX_MR = 0x00001000,
+ RXE_MAX_MW = 0x00001000,
RXE_MIN_MR_INDEX = 0x00000001,
- RXE_MAX_MR_INDEX = 0x00040000,
- RXE_MIN_MW_INDEX = 0x00040001,
- RXE_MAX_MW_INDEX = 0x00060000,
+ RXE_MAX_MR_INDEX = 0x00010000,
+ RXE_MIN_MW_INDEX = 0x00010001,
+ RXE_MAX_MW_INDEX = 0x00020000,
+
RXE_MAX_PKT_PER_ACK = 64,
RXE_MAX_UNACKED_PSNS = 128,
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index 11eba7a3ba8f..8d32e3f50813 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -273,7 +273,6 @@ enum rxe_mr_type {
RXE_MR_TYPE_NONE,
RXE_MR_TYPE_DMA,
RXE_MR_TYPE_MR,
- RXE_MR_TYPE_MW,
};
#define RXE_BUF_PER_MAP (PAGE_SIZE / sizeof(struct rxe_phys_buf))
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 4/9] RDMA/rxe: Add ib_alloc_mw and ib_dealloc_mw verbs
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (2 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 3/9] RDMA/rxe: Enable MW object pool Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 5/9] RDMA/rxe: Replace WR_REG_MASK by WR_LOCAL_OP_MASK Bob Pearson
` (5 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Add ib_alloc_mw and ib_dealloc_mw verbs APIs.
Added new file rxe_mw.c focused on MWs.
Changed the 8 bit random key generator.
Added a cleanup routine for MWs.
Added verbs routines to ib_device_ops.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/Makefile | 1 +
drivers/infiniband/sw/rxe/rxe_loc.h | 6 +++
drivers/infiniband/sw/rxe/rxe_mr.c | 20 +++++-----
drivers/infiniband/sw/rxe/rxe_mw.c | 53 +++++++++++++++++++++++++++
drivers/infiniband/sw/rxe/rxe_pool.c | 1 +
drivers/infiniband/sw/rxe/rxe_verbs.c | 3 ++
drivers/infiniband/sw/rxe/rxe_verbs.h | 2 +
7 files changed, 75 insertions(+), 11 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/Makefile b/drivers/infiniband/sw/rxe/Makefile
index 66af72dca759..1e24673e9318 100644
--- a/drivers/infiniband/sw/rxe/Makefile
+++ b/drivers/infiniband/sw/rxe/Makefile
@@ -15,6 +15,7 @@ rdma_rxe-y := \
rxe_qp.o \
rxe_cq.o \
rxe_mr.o \
+ rxe_mw.o \
rxe_opcode.o \
rxe_mmap.o \
rxe_icrc.o \
diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h
index ef8061d2fbe0..edf575930a98 100644
--- a/drivers/infiniband/sw/rxe/rxe_loc.h
+++ b/drivers/infiniband/sw/rxe/rxe_loc.h
@@ -76,6 +76,7 @@ enum copy_direction {
from_mr_obj,
};
+u8 rxe_get_next_key(u32 last_key);
void rxe_mr_init_dma(struct rxe_pd *pd, int access, struct rxe_mr *mr);
int rxe_mr_init_user(struct rxe_pd *pd, u64 start, u64 length, u64 iova,
@@ -106,6 +107,11 @@ void rxe_mr_cleanup(struct rxe_pool_entry *arg);
int advance_dma_data(struct rxe_dma_info *dma, unsigned int length);
+/* rxe_mw.c */
+int rxe_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata);
+int rxe_dealloc_mw(struct ib_mw *ibmw);
+void rxe_mw_cleanup(struct rxe_pool_entry *arg);
+
/* rxe_net.c */
void rxe_loopback(struct sk_buff *skb);
int rxe_send(struct rxe_pkt_info *pkt, struct sk_buff *skb);
diff --git a/drivers/infiniband/sw/rxe/rxe_mr.c b/drivers/infiniband/sw/rxe/rxe_mr.c
index 9f63947bab12..7f2cfc1ce659 100644
--- a/drivers/infiniband/sw/rxe/rxe_mr.c
+++ b/drivers/infiniband/sw/rxe/rxe_mr.c
@@ -7,19 +7,17 @@
#include "rxe.h"
#include "rxe_loc.h"
-/*
- * lfsr (linear feedback shift register) with period 255
+/* Return a random 8 bit key value that is
+ * different than the last_key. Set last_key to -1
+ * if this is the first key for an MR or MW
*/
-static u8 rxe_get_key(void)
+u8 rxe_get_next_key(u32 last_key)
{
- static u32 key = 1;
-
- key = key << 1;
-
- key |= (0 != (key & 0x100)) ^ (0 != (key & 0x10))
- ^ (0 != (key & 0x80)) ^ (0 != (key & 0x40));
+ u8 key;
- key &= 0xff;
+ do {
+ get_random_bytes(&key, 1);
+ } while (key == last_key);
return key;
}
@@ -47,7 +45,7 @@ int mr_check_range(struct rxe_mr *mr, u64 iova, size_t length)
static void rxe_mr_init(int access, struct rxe_mr *mr)
{
- u32 lkey = mr->pelem.index << 8 | rxe_get_key();
+ u32 lkey = mr->pelem.index << 8 | rxe_get_next_key(-1);
u32 rkey = (access & IB_ACCESS_REMOTE) ? lkey : 0;
mr->ibmr.lkey = lkey;
diff --git a/drivers/infiniband/sw/rxe/rxe_mw.c b/drivers/infiniband/sw/rxe/rxe_mw.c
new file mode 100644
index 000000000000..69128e298d44
--- /dev/null
+++ b/drivers/infiniband/sw/rxe/rxe_mw.c
@@ -0,0 +1,53 @@
+// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
+/*
+ * Copyright (c) 2020 Hewlett Packard Enterprise, Inc. All rights reserved.
+ */
+
+#include "rxe.h"
+
+int rxe_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
+{
+ struct rxe_mw *mw = to_rmw(ibmw);
+ struct rxe_pd *pd = to_rpd(ibmw->pd);
+ struct rxe_dev *rxe = to_rdev(ibmw->device);
+ int ret;
+
+ rxe_add_ref(pd);
+
+ ret = rxe_add_to_pool(&rxe->mw_pool, mw);
+ if (ret) {
+ rxe_drop_ref(pd);
+ return ret;
+ }
+
+ rxe_add_index(mw);
+ ibmw->rkey = (mw->pelem.index << 8) | rxe_get_next_key(-1);
+ mw->state = (mw->ibmw.type == IB_MW_TYPE_2) ?
+ RXE_MW_STATE_FREE : RXE_MW_STATE_VALID;
+ spin_lock_init(&mw->lock);
+
+ return 0;
+}
+
+int rxe_dealloc_mw(struct ib_mw *ibmw)
+{
+ struct rxe_mw *mw = to_rmw(ibmw);
+ struct rxe_pd *pd = to_rpd(ibmw->pd);
+ unsigned long flags;
+
+ spin_lock_irqsave(&mw->lock, flags);
+ mw->state = RXE_MW_STATE_INVALID;
+ spin_unlock_irqrestore(&mw->lock, flags);
+
+ rxe_drop_ref(mw);
+ rxe_drop_ref(pd);
+
+ return 0;
+}
+
+void rxe_mw_cleanup(struct rxe_pool_entry *elem)
+{
+ struct rxe_mw *mw = container_of(elem, typeof(*mw), pelem);
+
+ rxe_drop_index(mw);
+}
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index 2b795e2fc4b3..5b3277e8c35d 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -65,6 +65,7 @@ struct rxe_type_info rxe_type_info[RXE_NUM_TYPES] = {
.name = "rxe-mw",
.size = sizeof(struct rxe_mw),
.elem_offset = offsetof(struct rxe_mw, pelem),
+ .cleanup = rxe_mw_cleanup,
.flags = RXE_POOL_INDEX | RXE_POOL_NO_ALLOC,
.max_index = RXE_MAX_MW_INDEX,
.min_index = RXE_MIN_MW_INDEX,
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index aeb5e232c195..fff81bf78a86 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -1060,6 +1060,7 @@ static const struct ib_device_ops rxe_dev_ops = {
.alloc_hw_stats = rxe_ib_alloc_hw_stats,
.alloc_mr = rxe_alloc_mr,
+ .alloc_mw = rxe_alloc_mw,
.alloc_pd = rxe_alloc_pd,
.alloc_ucontext = rxe_alloc_ucontext,
.attach_mcast = rxe_attach_mcast,
@@ -1069,6 +1070,7 @@ static const struct ib_device_ops rxe_dev_ops = {
.create_srq = rxe_create_srq,
.create_user_ah = rxe_create_ah,
.dealloc_driver = rxe_dealloc,
+ .dealloc_mw = rxe_dealloc_mw,
.dealloc_pd = rxe_dealloc_pd,
.dealloc_ucontext = rxe_dealloc_ucontext,
.dereg_mr = rxe_dereg_mr,
@@ -1106,6 +1108,7 @@ static const struct ib_device_ops rxe_dev_ops = {
INIT_RDMA_OBJ_SIZE(ib_ah, rxe_ah, ibah),
INIT_RDMA_OBJ_SIZE(ib_cq, rxe_cq, ibcq),
+ INIT_RDMA_OBJ_SIZE(ib_mw, rxe_mw, ibmw),
INIT_RDMA_OBJ_SIZE(ib_pd, rxe_pd, ibpd),
INIT_RDMA_OBJ_SIZE(ib_srq, rxe_srq, ibsrq),
INIT_RDMA_OBJ_SIZE(ib_ucontext, rxe_ucontext, ibuc),
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index 8d32e3f50813..c8597ae8c833 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -323,6 +323,8 @@ enum rxe_mw_state {
struct rxe_mw {
struct ib_mw ibmw;
struct rxe_pool_entry pelem;
+ spinlock_t lock;
+ enum rxe_mw_state state;
};
struct rxe_mc_grp {
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 5/9] RDMA/rxe: Replace WR_REG_MASK by WR_LOCAL_OP_MASK
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (3 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 4/9] RDMA/rxe: Add ib_alloc_mw and ib_dealloc_mw verbs Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 6/9] RDMA/rxe: Move local ops to subroutine Bob Pearson
` (4 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Rxe has two mask bits WR_LOCAL_MASK and WR_REG_MASK with
WR_REG_MASK used to indicate any local operation and WR_LOCAL_MASK
unused. This patch replaces both of these with one mask bit
WR_LOCAL_OP_MASK which is clearer.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe_opcode.c | 4 ++--
drivers/infiniband/sw/rxe/rxe_opcode.h | 3 +--
drivers/infiniband/sw/rxe/rxe_req.c | 2 +-
drivers/infiniband/sw/rxe/rxe_verbs.c | 2 +-
4 files changed, 5 insertions(+), 6 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_opcode.c b/drivers/infiniband/sw/rxe/rxe_opcode.c
index 0cb4b01fd910..1e4b67b048f3 100644
--- a/drivers/infiniband/sw/rxe/rxe_opcode.c
+++ b/drivers/infiniband/sw/rxe/rxe_opcode.c
@@ -87,13 +87,13 @@ struct rxe_wr_opcode_info rxe_wr_opcode_info[] = {
[IB_WR_LOCAL_INV] = {
.name = "IB_WR_LOCAL_INV",
.mask = {
- [IB_QPT_RC] = WR_REG_MASK,
+ [IB_QPT_RC] = WR_LOCAL_OP_MASK,
},
},
[IB_WR_REG_MR] = {
.name = "IB_WR_REG_MR",
.mask = {
- [IB_QPT_RC] = WR_REG_MASK,
+ [IB_QPT_RC] = WR_LOCAL_OP_MASK,
},
},
};
diff --git a/drivers/infiniband/sw/rxe/rxe_opcode.h b/drivers/infiniband/sw/rxe/rxe_opcode.h
index 1041ac9a9233..e02f039b8c44 100644
--- a/drivers/infiniband/sw/rxe/rxe_opcode.h
+++ b/drivers/infiniband/sw/rxe/rxe_opcode.h
@@ -19,8 +19,7 @@ enum rxe_wr_mask {
WR_SEND_MASK = BIT(2),
WR_READ_MASK = BIT(3),
WR_WRITE_MASK = BIT(4),
- WR_LOCAL_MASK = BIT(5),
- WR_REG_MASK = BIT(6),
+ WR_LOCAL_OP_MASK = BIT(5),
WR_READ_OR_WRITE_MASK = WR_READ_MASK | WR_WRITE_MASK,
WR_READ_WRITE_OR_SEND_MASK = WR_READ_OR_WRITE_MASK | WR_SEND_MASK,
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 3664cdae7e1f..0d4dcd514c55 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -593,7 +593,7 @@ int rxe_requester(void *arg)
if (unlikely(!wqe))
goto exit;
- if (wqe->mask & WR_REG_MASK) {
+ if (wqe->mask & WR_LOCAL_OP_MASK) {
if (wqe->wr.opcode == IB_WR_LOCAL_INV) {
struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
struct rxe_mr *rmr;
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index fff81bf78a86..d22f011a20f3 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -577,7 +577,7 @@ static void init_send_wqe(struct rxe_qp *qp, const struct ib_send_wr *ibwr,
init_send_wr(qp, &wqe->wr, ibwr);
/* local operation */
- if (unlikely(mask & WR_REG_MASK)) {
+ if (unlikely(mask & WR_LOCAL_OP_MASK)) {
wqe->mask = mask;
wqe->state = wqe_state_posted;
return;
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 6/9] RDMA/rxe: Move local ops to subroutine
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (4 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 5/9] RDMA/rxe: Replace WR_REG_MASK by WR_LOCAL_OP_MASK Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 7/9] RDMA/rxe: Add support for bind MW work requests Bob Pearson
` (3 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Simplify rxe_requester() by moving the local operations
to a subroutine. Add an error return for illegal send WR opcode.
Moved next_index ahead of rxe_run_task which fixed a small bug where
work completions were delayed until after the next wqe which was not
the intended behavior.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe_req.c | 89 +++++++++++++++++------------
1 file changed, 54 insertions(+), 35 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 0d4dcd514c55..0cf97e3db29f 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -555,6 +555,56 @@ static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
jiffies + qp->qp_timeout_jiffies);
}
+static int do_local_ops(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
+{
+ u8 opcode = wqe->wr.opcode;
+ struct rxe_dev *rxe;
+ struct rxe_mr *mr;
+ u32 rkey;
+
+ switch (opcode) {
+ case IB_WR_LOCAL_INV:
+ rxe = to_rdev(qp->ibqp.device);
+ rkey = wqe->wr.ex.invalidate_rkey;
+ mr = rxe_pool_get_index(&rxe->mr_pool, rkey >> 8);
+ if (!mr) {
+ pr_err("No MR for rkey %#x\n", rkey);
+ wqe->state = wqe_state_error;
+ wqe->status = IB_WC_LOC_QP_OP_ERR;
+ return -EINVAL;
+ }
+ mr->state = RXE_MR_STATE_FREE;
+ rxe_drop_ref(mr);
+ break;
+ case IB_WR_REG_MR:
+ mr = to_rmr(wqe->wr.wr.reg.mr);
+
+ rxe_add_ref(mr);
+ mr->state = RXE_MR_STATE_VALID;
+ mr->access = wqe->wr.wr.reg.access;
+ mr->ibmr.lkey = wqe->wr.wr.reg.key;
+ mr->ibmr.rkey = wqe->wr.wr.reg.key;
+ mr->iova = wqe->wr.wr.reg.mr->iova;
+ rxe_drop_ref(mr);
+ break;
+ default:
+ pr_err("Unexpected send wqe opcode %d\n", opcode);
+ wqe->state = wqe_state_error;
+ wqe->status = IB_WC_LOC_QP_OP_ERR;
+ return -EINVAL;
+ }
+
+ wqe->state = wqe_state_done;
+ wqe->status = IB_WC_SUCCESS;
+ qp->req.wqe_index = next_index(qp->sq.queue, qp->req.wqe_index);
+
+ if ((wqe->wr.send_flags & IB_SEND_SIGNALED) ||
+ qp->sq_sig_type == IB_SIGNAL_ALL_WR)
+ rxe_run_task(&qp->comp.task, 1);
+
+ return 0;
+}
+
int rxe_requester(void *arg)
{
struct rxe_qp *qp = (struct rxe_qp *)arg;
@@ -594,42 +644,11 @@ int rxe_requester(void *arg)
goto exit;
if (wqe->mask & WR_LOCAL_OP_MASK) {
- if (wqe->wr.opcode == IB_WR_LOCAL_INV) {
- struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
- struct rxe_mr *rmr;
-
- rmr = rxe_pool_get_index(&rxe->mr_pool,
- wqe->wr.ex.invalidate_rkey >> 8);
- if (!rmr) {
- pr_err("No mr for key %#x\n",
- wqe->wr.ex.invalidate_rkey);
- wqe->state = wqe_state_error;
- wqe->status = IB_WC_MW_BIND_ERR;
- goto exit;
- }
- rmr->state = RXE_MR_STATE_FREE;
- rxe_drop_ref(rmr);
- wqe->state = wqe_state_done;
- wqe->status = IB_WC_SUCCESS;
- } else if (wqe->wr.opcode == IB_WR_REG_MR) {
- struct rxe_mr *rmr = to_rmr(wqe->wr.wr.reg.mr);
-
- rmr->state = RXE_MR_STATE_VALID;
- rmr->access = wqe->wr.wr.reg.access;
- rmr->ibmr.lkey = wqe->wr.wr.reg.key;
- rmr->ibmr.rkey = wqe->wr.wr.reg.key;
- rmr->iova = wqe->wr.wr.reg.mr->iova;
- wqe->state = wqe_state_done;
- wqe->status = IB_WC_SUCCESS;
- } else {
+ ret = do_local_ops(qp, wqe);
+ if (unlikely(ret))
goto exit;
- }
- if ((wqe->wr.send_flags & IB_SEND_SIGNALED) ||
- qp->sq_sig_type == IB_SIGNAL_ALL_WR)
- rxe_run_task(&qp->comp.task, 1);
- qp->req.wqe_index = next_index(qp->sq.queue,
- qp->req.wqe_index);
- goto next_wqe;
+ else
+ goto next_wqe;
}
if (unlikely(qp_type(qp) == IB_QPT_RC &&
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 7/9] RDMA/rxe: Add support for bind MW work requests
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (5 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 6/9] RDMA/rxe: Move local ops to subroutine Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 8/9] RDMA/rxe: Implement invalidate MW operations Bob Pearson
` (2 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Add support for bind MW work requests from user space.
Since rdma/core does not support bind mw in ib_send_wr
there is no way to easily support bind mw in kernel space.
Added bind_mw local operation in rxe_req.c
Added bind_mw WR operation in rxe_opcode.c
Added bind_mw WC in rxe_comp.c
Added additional fields to rxe_mw in rxe_verbs.h
Added do_dealloc_mw() subroutine to cleanup an mw
when rxe_dealloc_mw is called.
Added code to implement bind_mw operation in rxe_mw.c
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe_comp.c | 1 +
drivers/infiniband/sw/rxe/rxe_loc.h | 1 +
drivers/infiniband/sw/rxe/rxe_mw.c | 215 ++++++++++++++++++++++++-
drivers/infiniband/sw/rxe/rxe_opcode.c | 7 +
drivers/infiniband/sw/rxe/rxe_req.c | 9 ++
drivers/infiniband/sw/rxe/rxe_verbs.h | 15 +-
6 files changed, 243 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_comp.c b/drivers/infiniband/sw/rxe/rxe_comp.c
index a612b335baa0..28263338c484 100644
--- a/drivers/infiniband/sw/rxe/rxe_comp.c
+++ b/drivers/infiniband/sw/rxe/rxe_comp.c
@@ -103,6 +103,7 @@ static enum ib_wc_opcode wr_to_wc_opcode(enum ib_wr_opcode opcode)
case IB_WR_RDMA_READ_WITH_INV: return IB_WC_RDMA_READ;
case IB_WR_LOCAL_INV: return IB_WC_LOCAL_INV;
case IB_WR_REG_MR: return IB_WC_REG_MR;
+ case IB_WR_BIND_MW: return IB_WC_BIND_MW;
default:
return 0xff;
diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h
index edf575930a98..e6f574973298 100644
--- a/drivers/infiniband/sw/rxe/rxe_loc.h
+++ b/drivers/infiniband/sw/rxe/rxe_loc.h
@@ -110,6 +110,7 @@ int advance_dma_data(struct rxe_dma_info *dma, unsigned int length);
/* rxe_mw.c */
int rxe_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata);
int rxe_dealloc_mw(struct ib_mw *ibmw);
+int rxe_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe);
void rxe_mw_cleanup(struct rxe_pool_entry *arg);
/* rxe_net.c */
diff --git a/drivers/infiniband/sw/rxe/rxe_mw.c b/drivers/infiniband/sw/rxe/rxe_mw.c
index 69128e298d44..946d6e85b6e9 100644
--- a/drivers/infiniband/sw/rxe/rxe_mw.c
+++ b/drivers/infiniband/sw/rxe/rxe_mw.c
@@ -29,6 +29,29 @@ int rxe_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
return 0;
}
+static void do_dealloc_mw(struct rxe_mw *mw)
+{
+ if (mw->mr) {
+ struct rxe_mr *mr = mw->mr;
+
+ mw->mr = NULL;
+ atomic_dec(&mr->num_mw);
+ rxe_drop_ref(mr);
+ }
+
+ if (mw->qp) {
+ struct rxe_qp *qp = mw->qp;
+
+ mw->qp = NULL;
+ rxe_drop_ref(qp);
+ }
+
+ mw->access = 0;
+ mw->addr = 0;
+ mw->length = 0;
+ mw->state = RXE_MW_STATE_INVALID;
+}
+
int rxe_dealloc_mw(struct ib_mw *ibmw)
{
struct rxe_mw *mw = to_rmw(ibmw);
@@ -36,7 +59,7 @@ int rxe_dealloc_mw(struct ib_mw *ibmw)
unsigned long flags;
spin_lock_irqsave(&mw->lock, flags);
- mw->state = RXE_MW_STATE_INVALID;
+ do_dealloc_mw(mw);
spin_unlock_irqrestore(&mw->lock, flags);
rxe_drop_ref(mw);
@@ -45,6 +68,196 @@ int rxe_dealloc_mw(struct ib_mw *ibmw)
return 0;
}
+static int check_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
+ struct rxe_mw *mw, struct rxe_mr *mr)
+{
+ if (mw->ibmw.type == IB_MW_TYPE_1) {
+ if (unlikely(mw->state != RXE_MW_STATE_VALID)) {
+ pr_err_once(
+ "attempt to bind a type 1 MW not in the valid state\n");
+ return -EINVAL;
+ }
+
+ /* o10-36.2.2 */
+ if (unlikely((mw->access & IB_ZERO_BASED))) {
+ pr_err_once("attempt to bind a zero based type 1 MW\n");
+ return -EINVAL;
+ }
+ }
+
+ if (mw->ibmw.type == IB_MW_TYPE_2) {
+ /* o10-37.2.30 */
+ if (unlikely(mw->state != RXE_MW_STATE_FREE)) {
+ pr_err_once(
+ "attempt to bind a type 2 MW not in the free state\n");
+ return -EINVAL;
+ }
+
+ /* C10-72 */
+ if (unlikely(qp->pd != to_rpd(mw->ibmw.pd))) {
+ pr_err_once(
+ "attempt to bind type 2 MW with qp with different PD\n");
+ return -EINVAL;
+ }
+
+ /* o10-37.2.40 */
+ if (unlikely(!mr || wqe->wr.wr.umw.length == 0)) {
+ pr_err_once(
+ "attempt to invalidate type 2 MW by binding with NULL or zero length MR\n");
+ return -EINVAL;
+ }
+ }
+
+ if (unlikely((wqe->wr.wr.umw.rkey & 0xff) == (mw->ibmw.rkey & 0xff))) {
+ pr_err_once("attempt to bind MW with same key\n");
+ return -EINVAL;
+ }
+
+ /* remaining checks only apply to a nonzero MR */
+ if (!mr)
+ return 0;
+
+ if (unlikely(mr->access & IB_ZERO_BASED)) {
+ pr_err_once("attempt to bind MW to zero based MR\n");
+ return -EINVAL;
+ }
+
+ /* C10-73 */
+ if (unlikely(!(mr->access & IB_ACCESS_MW_BIND))) {
+ pr_err_once(
+ "attempt to bind an MW to an MR without bind access\n");
+ return -EINVAL;
+ }
+
+ /* C10-74 */
+ if (unlikely((mw->access &
+ (IB_ACCESS_REMOTE_WRITE | IB_ACCESS_REMOTE_ATOMIC)) &&
+ !(mr->access & IB_ACCESS_LOCAL_WRITE))) {
+ pr_err_once(
+ "attempt to bind an writeable MW to an MR without local write access\n");
+ return -EINVAL;
+ }
+
+ /* C10-75 */
+ if (mw->access & IB_ZERO_BASED) {
+ if (unlikely(wqe->wr.wr.umw.length > mr->length)) {
+ pr_err_once(
+ "attempt to bind a ZB MW outside of the MR\n");
+ return -EINVAL;
+ }
+ } else {
+ if (unlikely((wqe->wr.wr.umw.addr < mr->iova) ||
+ ((wqe->wr.wr.umw.addr + wqe->wr.wr.umw.length) >
+ (mr->iova + mr->length)))) {
+ pr_err_once(
+ "attempt to bind a VA MW outside of the MR\n");
+ return -EINVAL;
+ }
+ }
+
+ return 0;
+}
+
+static int do_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
+ struct rxe_mw *mw, struct rxe_mr *mr)
+{
+ u32 rkey;
+ u32 new_rkey;
+
+ rkey = mw->ibmw.rkey;
+ new_rkey = (rkey & 0xffffff00) | (wqe->wr.wr.umw.rkey & 0x000000ff);
+
+ mw->ibmw.rkey = new_rkey;
+ mw->access = wqe->wr.wr.umw.access;
+ mw->state = RXE_MW_STATE_VALID;
+ mw->addr = wqe->wr.wr.umw.addr;
+ mw->length = wqe->wr.wr.umw.length;
+
+ if (mw->mr) {
+ rxe_drop_ref(mw->mr);
+ atomic_dec(&mw->mr->num_mw);
+ mw->mr = NULL;
+ }
+
+ if (mw->length) {
+ mw->mr = mr;
+ atomic_inc(&mr->num_mw);
+ rxe_add_ref(mr);
+ }
+
+ if (mw->ibmw.type == IB_MW_TYPE_2) {
+ rxe_add_ref(qp);
+ mw->qp = qp;
+ }
+
+ return 0;
+}
+
+int rxe_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
+{
+ int ret;
+ struct rxe_mw *mw;
+ struct rxe_mr *mr;
+ struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
+ unsigned long flags;
+
+ if (qp->is_user) {
+ mw = rxe_pool_get_index(&rxe->mw_pool,
+ wqe->wr.wr.umw.mw_rkey >> 8);
+ if (unlikely(!mw)) {
+ ret = -EINVAL;
+ goto err;
+ }
+
+ if (unlikely(mw->ibmw.rkey != wqe->wr.wr.umw.mw_rkey)) {
+ ret = -EINVAL;
+ goto err_drop_mw;
+ }
+
+ if (likely(wqe->wr.wr.umw.length)) {
+ mr = rxe_pool_get_index(&rxe->mr_pool,
+ wqe->wr.wr.umw.mr_lkey >> 8);
+ if (unlikely(!mr)) {
+ ret = -EINVAL;
+ goto err_drop_mw;
+ }
+
+ if (unlikely(mr->ibmr.lkey != wqe->wr.wr.umw.mr_lkey)) {
+ ret = -EINVAL;
+ goto err_drop_mr;
+ }
+ } else {
+ mr = NULL;
+ }
+ } else {
+ mw = to_rmw(wqe->wr.wr.kmw.mw);
+ rxe_add_ref(mw);
+ if (wqe->wr.wr.kmw.mr) {
+ mr = to_rmr(wqe->wr.wr.kmw.mr);
+ rxe_add_ref(mr);
+ } else {
+ mr = NULL;
+ }
+ }
+
+ spin_lock_irqsave(&mw->lock, flags);
+
+ ret = check_bind_mw(qp, wqe, mw, mr);
+ if (ret)
+ goto err_unlock;
+
+ ret = do_bind_mw(qp, wqe, mw, mr);
+err_unlock:
+ spin_unlock_irqrestore(&mw->lock, flags);
+err_drop_mr:
+ if (mr)
+ rxe_drop_ref(mr);
+err_drop_mw:
+ rxe_drop_ref(mw);
+err:
+ return ret;
+}
+
void rxe_mw_cleanup(struct rxe_pool_entry *elem)
{
struct rxe_mw *mw = container_of(elem, typeof(*mw), pelem);
diff --git a/drivers/infiniband/sw/rxe/rxe_opcode.c b/drivers/infiniband/sw/rxe/rxe_opcode.c
index 1e4b67b048f3..3ef5a10a6efd 100644
--- a/drivers/infiniband/sw/rxe/rxe_opcode.c
+++ b/drivers/infiniband/sw/rxe/rxe_opcode.c
@@ -96,6 +96,13 @@ struct rxe_wr_opcode_info rxe_wr_opcode_info[] = {
[IB_QPT_RC] = WR_LOCAL_OP_MASK,
},
},
+ [IB_WR_BIND_MW] = {
+ .name = "IB_WR_BIND_MW",
+ .mask = {
+ [IB_QPT_RC] = WR_LOCAL_OP_MASK,
+ [IB_QPT_UC] = WR_LOCAL_OP_MASK,
+ },
+ },
};
struct rxe_opcode_info rxe_opcode[RXE_NUM_OPCODE] = {
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 0cf97e3db29f..243602584a28 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -561,6 +561,7 @@ static int do_local_ops(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
struct rxe_dev *rxe;
struct rxe_mr *mr;
u32 rkey;
+ int ret;
switch (opcode) {
case IB_WR_LOCAL_INV:
@@ -587,6 +588,14 @@ static int do_local_ops(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
mr->iova = wqe->wr.wr.reg.mr->iova;
rxe_drop_ref(mr);
break;
+ case IB_WR_BIND_MW:
+ ret = rxe_bind_mw(qp, wqe);
+ if (ret) {
+ wqe->state = wqe_state_error;
+ wqe->status = IB_WC_MW_BIND_ERR;
+ return -EINVAL;
+ }
+ break;
default:
pr_err("Unexpected send wqe opcode %d\n", opcode);
wqe->state = wqe_state_error;
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index c8597ae8c833..7da47b8c707b 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -312,6 +312,8 @@ struct rxe_mr {
u32 num_map;
struct rxe_map **map;
+
+ atomic_t num_mw;
};
enum rxe_mw_state {
@@ -321,10 +323,15 @@ enum rxe_mw_state {
};
struct rxe_mw {
- struct ib_mw ibmw;
- struct rxe_pool_entry pelem;
- spinlock_t lock;
- enum rxe_mw_state state;
+ struct ib_mw ibmw;
+ struct rxe_pool_entry pelem;
+ spinlock_t lock;
+ enum rxe_mw_state state;
+ struct rxe_qp *qp; /* Type 2 only */
+ struct rxe_mr *mr;
+ int access;
+ u64 addr;
+ u64 length;
};
struct rxe_mc_grp {
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 8/9] RDMA/rxe: Implement invalidate MW operations
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (6 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 7/9] RDMA/rxe: Add support for bind MW work requests Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-08 21:40 ` [PATCH for-next 9/9] RDMA/rxe: Implement memory access through MWs Bob Pearson
2021-04-13 23:22 ` [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Jason Gunthorpe
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Implement invalidate MW and cleaned up invalidate MR operations.
Added code to perform remote invalidate for send with invalidate.
Added code to perform local invalidation.
Deleted some blank lines in rxe_loc.h.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe_loc.h | 23 ++++-----
drivers/infiniband/sw/rxe/rxe_mr.c | 59 +++++++++++++++++------
drivers/infiniband/sw/rxe/rxe_mw.c | 67 +++++++++++++++++++++++++++
drivers/infiniband/sw/rxe/rxe_req.c | 22 +++++----
drivers/infiniband/sw/rxe/rxe_resp.c | 52 +++++++++++++--------
drivers/infiniband/sw/rxe/rxe_verbs.h | 23 +++++----
6 files changed, 178 insertions(+), 68 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h
index e6f574973298..7f1117c51e30 100644
--- a/drivers/infiniband/sw/rxe/rxe_loc.h
+++ b/drivers/infiniband/sw/rxe/rxe_loc.h
@@ -76,41 +76,34 @@ enum copy_direction {
from_mr_obj,
};
+enum lookup_type {
+ lookup_local,
+ lookup_remote,
+};
+
u8 rxe_get_next_key(u32 last_key);
void rxe_mr_init_dma(struct rxe_pd *pd, int access, struct rxe_mr *mr);
-
int rxe_mr_init_user(struct rxe_pd *pd, u64 start, u64 length, u64 iova,
int access, struct ib_udata *udata, struct rxe_mr *mr);
-
int rxe_mr_init_fast(struct rxe_pd *pd, int max_pages, struct rxe_mr *mr);
-
int rxe_mr_copy(struct rxe_mr *mr, u64 iova, void *addr, int length,
enum copy_direction dir, u32 *crcp);
-
int copy_data(struct rxe_pd *pd, int access,
struct rxe_dma_info *dma, void *addr, int length,
enum copy_direction dir, u32 *crcp);
-
void *iova_to_vaddr(struct rxe_mr *mr, u64 iova, int length);
-
-enum lookup_type {
- lookup_local,
- lookup_remote,
-};
-
struct rxe_mr *lookup_mr(struct rxe_pd *pd, int access, u32 key,
enum lookup_type type);
-
int mr_check_range(struct rxe_mr *mr, u64 iova, size_t length);
-
-void rxe_mr_cleanup(struct rxe_pool_entry *arg);
-
int advance_dma_data(struct rxe_dma_info *dma, unsigned int length);
+int rxe_invalidate_mr(struct rxe_qp *qp, u32 rkey);
+void rxe_mr_cleanup(struct rxe_pool_entry *arg);
/* rxe_mw.c */
int rxe_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata);
int rxe_dealloc_mw(struct ib_mw *ibmw);
int rxe_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe);
+int rxe_invalidate_mw(struct rxe_qp *qp, u32 rkey);
void rxe_mw_cleanup(struct rxe_pool_entry *arg);
/* rxe_net.c */
diff --git a/drivers/infiniband/sw/rxe/rxe_mr.c b/drivers/infiniband/sw/rxe/rxe_mr.c
index 7f2cfc1ce659..0f1791ed0350 100644
--- a/drivers/infiniband/sw/rxe/rxe_mr.c
+++ b/drivers/infiniband/sw/rxe/rxe_mr.c
@@ -55,21 +55,6 @@ static void rxe_mr_init(int access, struct rxe_mr *mr)
mr->map_shift = ilog2(RXE_BUF_PER_MAP);
}
-void rxe_mr_cleanup(struct rxe_pool_entry *arg)
-{
- struct rxe_mr *mr = container_of(arg, typeof(*mr), pelem);
- int i;
-
- ib_umem_release(mr->umem);
-
- if (mr->map) {
- for (i = 0; i < mr->num_map; i++)
- kfree(mr->map[i]);
-
- kfree(mr->map);
- }
-}
-
static int rxe_mr_alloc(struct rxe_mr *mr, int num_buf)
{
int i;
@@ -540,3 +525,47 @@ struct rxe_mr *lookup_mr(struct rxe_pd *pd, int access, u32 key,
return mr;
}
+
+int rxe_invalidate_mr(struct rxe_qp *qp, u32 rkey)
+{
+ struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
+ struct rxe_mr *mr;
+ int ret;
+
+ mr = rxe_pool_get_index(&rxe->mr_pool, rkey >> 8);
+ if (!mr) {
+ pr_err("%s: No MR for rkey %#x\n", __func__, rkey);
+ ret = -EINVAL;
+ goto err;
+ }
+
+ if (rkey != mr->ibmr.rkey) {
+ pr_err("%s: rkey (%#x) doesn't match mr->ibmr.rkey (%#x)\n",
+ __func__, rkey, mr->ibmr.rkey);
+ ret = -EINVAL;
+ goto err_drop_ref;
+ }
+
+ mr->state = RXE_MR_STATE_FREE;
+ ret = 0;
+
+err_drop_ref:
+ rxe_drop_ref(mr);
+err:
+ return ret;
+}
+
+void rxe_mr_cleanup(struct rxe_pool_entry *arg)
+{
+ struct rxe_mr *mr = container_of(arg, typeof(*mr), pelem);
+ int i;
+
+ ib_umem_release(mr->umem);
+
+ if (mr->map) {
+ for (i = 0; i < mr->num_map; i++)
+ kfree(mr->map[i]);
+
+ kfree(mr->map);
+ }
+}
diff --git a/drivers/infiniband/sw/rxe/rxe_mw.c b/drivers/infiniband/sw/rxe/rxe_mw.c
index 946d6e85b6e9..7b39670f32cd 100644
--- a/drivers/infiniband/sw/rxe/rxe_mw.c
+++ b/drivers/infiniband/sw/rxe/rxe_mw.c
@@ -258,6 +258,73 @@ int rxe_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
return ret;
}
+static int check_invalidate_mw(struct rxe_qp *qp, struct rxe_mw *mw)
+{
+ if (unlikely(mw->state == RXE_MW_STATE_INVALID))
+ return -EINVAL;
+
+ /* o10-37.2.26 */
+ if (unlikely(mw->ibmw.type == IB_MW_TYPE_1))
+ return -EINVAL;
+
+ return 0;
+}
+
+static void do_invalidate_mw(struct rxe_mw *mw)
+{
+ struct rxe_qp *qp;
+ struct rxe_mr *mr;
+
+ /* valid type 2 MW will always have a QP pointer */
+ qp = mw->qp;
+ mw->qp = NULL;
+ rxe_drop_ref(qp);
+
+ /* valid type 2 MW will always have an MR pointer */
+ mr = mw->mr;
+ mw->mr = NULL;
+ atomic_dec(&mr->num_mw);
+ rxe_drop_ref(mr);
+
+ mw->access = 0;
+ mw->addr = 0;
+ mw->length = 0;
+ mw->state = RXE_MW_STATE_FREE;
+}
+
+int rxe_invalidate_mw(struct rxe_qp *qp, u32 rkey)
+{
+ struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
+ unsigned long flags;
+ struct rxe_mw *mw;
+ int ret;
+
+ mw = rxe_pool_get_index(&rxe->mw_pool, rkey >> 8);
+ if (!mw) {
+ ret = -EINVAL;
+ goto err;
+ }
+
+ if (rkey != mw->ibmw.rkey) {
+ ret = -EINVAL;
+ goto err_drop_ref;
+ }
+
+ spin_lock_irqsave(&mw->lock, flags);
+
+ ret = check_invalidate_mw(qp, mw);
+ if (ret)
+ goto err_unlock;
+
+ do_invalidate_mw(mw);
+err_unlock:
+ spin_unlock_irqrestore(&mw->lock, flags);
+err_drop_ref:
+ rxe_drop_ref(mw);
+err:
+ return ret;
+}
+
void rxe_mw_cleanup(struct rxe_pool_entry *elem)
{
struct rxe_mw *mw = container_of(elem, typeof(*mw), pelem);
diff --git a/drivers/infiniband/sw/rxe/rxe_req.c b/drivers/infiniband/sw/rxe/rxe_req.c
index 243602584a28..66fc208d0ec1 100644
--- a/drivers/infiniband/sw/rxe/rxe_req.c
+++ b/drivers/infiniband/sw/rxe/rxe_req.c
@@ -558,25 +558,25 @@ static void update_state(struct rxe_qp *qp, struct rxe_send_wqe *wqe,
static int do_local_ops(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
{
u8 opcode = wqe->wr.opcode;
- struct rxe_dev *rxe;
struct rxe_mr *mr;
- u32 rkey;
int ret;
+ u32 rkey;
switch (opcode) {
case IB_WR_LOCAL_INV:
- rxe = to_rdev(qp->ibqp.device);
rkey = wqe->wr.ex.invalidate_rkey;
- mr = rxe_pool_get_index(&rxe->mr_pool, rkey >> 8);
- if (!mr) {
- pr_err("No MR for rkey %#x\n", rkey);
+ if (rkey_is_mw(rkey))
+ ret = rxe_invalidate_mw(qp, rkey);
+ else
+ ret = rxe_invalidate_mr(qp, rkey);
+
+ if (ret) {
wqe->state = wqe_state_error;
wqe->status = IB_WC_LOC_QP_OP_ERR;
- return -EINVAL;
+ return ret;
}
- mr->state = RXE_MR_STATE_FREE;
- rxe_drop_ref(mr);
break;
+
case IB_WR_REG_MR:
mr = to_rmr(wqe->wr.wr.reg.mr);
@@ -588,14 +588,16 @@ static int do_local_ops(struct rxe_qp *qp, struct rxe_send_wqe *wqe)
mr->iova = wqe->wr.wr.reg.mr->iova;
rxe_drop_ref(mr);
break;
+
case IB_WR_BIND_MW:
ret = rxe_bind_mw(qp, wqe);
if (ret) {
wqe->state = wqe_state_error;
wqe->status = IB_WC_MW_BIND_ERR;
- return -EINVAL;
+ return ret;
}
break;
+
default:
pr_err("Unexpected send wqe opcode %d\n", opcode);
wqe->state = wqe_state_error;
diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index 8e237b623b31..8834f8a23644 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -35,6 +35,7 @@ enum resp_states {
RESPST_ERR_TOO_MANY_RDMA_ATM_REQ,
RESPST_ERR_RNR,
RESPST_ERR_RKEY_VIOLATION,
+ RESPST_ERR_INVALIDATE_RKEY,
RESPST_ERR_LENGTH,
RESPST_ERR_CQ_OVERFLOW,
RESPST_ERROR,
@@ -68,6 +69,7 @@ static char *resp_state_name[] = {
[RESPST_ERR_TOO_MANY_RDMA_ATM_REQ] = "ERR_TOO_MANY_RDMA_ATM_REQ",
[RESPST_ERR_RNR] = "ERR_RNR",
[RESPST_ERR_RKEY_VIOLATION] = "ERR_RKEY_VIOLATION",
+ [RESPST_ERR_INVALIDATE_RKEY] = "ERR_INVALIDATE_RKEY_VIOLATION",
[RESPST_ERR_LENGTH] = "ERR_LENGTH",
[RESPST_ERR_CQ_OVERFLOW] = "ERR_CQ_OVERFLOW",
[RESPST_ERROR] = "ERROR",
@@ -751,6 +753,14 @@ static void build_rdma_network_hdr(union rdma_network_hdr *hdr,
memcpy(&hdr->ibgrh, ipv6_hdr(skb), sizeof(hdr->ibgrh));
}
+static int invalidate_rkey(struct rxe_qp *qp, u32 rkey)
+{
+ if (rkey_is_mw(rkey))
+ return rxe_invalidate_mw(qp, rkey);
+ else
+ return rxe_invalidate_mr(qp, rkey);
+}
+
/* Executes a new request. A retried request never reach that function (send
* and writes are discarded, and reads and atomics are retried elsewhere.
*/
@@ -790,6 +800,14 @@ static enum resp_states execute(struct rxe_qp *qp, struct rxe_pkt_info *pkt)
WARN_ON_ONCE(1);
}
+ if (pkt->mask & RXE_IETH_MASK) {
+ u32 rkey = ieth_rkey(pkt);
+
+ err = invalidate_rkey(qp, rkey);
+ if (err)
+ return RESPST_ERR_INVALIDATE_RKEY;
+ }
+
/* next expected psn, read handles this separately */
qp->resp.psn = (pkt->psn + 1) & BTH_PSN_MASK;
qp->resp.ack_psn = qp->resp.psn;
@@ -822,13 +840,13 @@ static enum resp_states do_complete(struct rxe_qp *qp,
memset(&cqe, 0, sizeof(cqe));
if (qp->rcq->is_user) {
- uwc->status = qp->resp.status;
- uwc->qp_num = qp->ibqp.qp_num;
- uwc->wr_id = wqe->wr_id;
+ uwc->status = qp->resp.status;
+ uwc->qp_num = qp->ibqp.qp_num;
+ uwc->wr_id = wqe->wr_id;
} else {
- wc->status = qp->resp.status;
- wc->qp = &qp->ibqp;
- wc->wr_id = wqe->wr_id;
+ wc->status = qp->resp.status;
+ wc->qp = &qp->ibqp;
+ wc->wr_id = wqe->wr_id;
}
if (wc->status == IB_WC_SUCCESS) {
@@ -883,27 +901,14 @@ static enum resp_states do_complete(struct rxe_qp *qp,
}
if (pkt->mask & RXE_IETH_MASK) {
- struct rxe_mr *rmr;
-
wc->wc_flags |= IB_WC_WITH_INVALIDATE;
wc->ex.invalidate_rkey = ieth_rkey(pkt);
-
- rmr = rxe_pool_get_index(&rxe->mr_pool,
- wc->ex.invalidate_rkey >> 8);
- if (unlikely(!rmr)) {
- pr_err("Bad rkey %#x invalidation\n",
- wc->ex.invalidate_rkey);
- return RESPST_ERROR;
- }
- rmr->state = RXE_MR_STATE_FREE;
- rxe_drop_ref(rmr);
}
- wc->qp = &qp->ibqp;
-
if (pkt->mask & RXE_DETH_MASK)
wc->src_qp = deth_sqp(pkt);
+ wc->qp = &qp->ibqp;
wc->port_num = qp->attr.port_num;
}
}
@@ -1316,6 +1321,13 @@ int rxe_responder(void *arg)
}
break;
+ case RESPST_ERR_INVALIDATE_RKEY:
+ /* RC - Class J. */
+ qp->resp.goto_error = 1;
+ qp->resp.status = IB_WC_REM_INV_REQ_ERR;
+ state = RESPST_COMPLETE;
+ break;
+
case RESPST_ERR_LENGTH:
if (qp_type(qp) == IB_QPT_RC) {
/* Class C */
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index 7da47b8c707b..b286a14ec282 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -286,6 +286,13 @@ struct rxe_map {
struct rxe_phys_buf buf[RXE_BUF_PER_MAP];
};
+static inline int rkey_is_mw(u32 rkey)
+{
+ u32 index = rkey >> 8;
+
+ return (index >= RXE_MIN_MW_INDEX) && (index <= RXE_MAX_MW_INDEX);
+}
+
struct rxe_mr {
struct rxe_pool_entry pelem;
struct ib_mr ibmr;
@@ -311,23 +318,23 @@ struct rxe_mr {
u32 max_buf;
u32 num_map;
- struct rxe_map **map;
-
atomic_t num_mw;
+
+ struct rxe_map **map;
};
enum rxe_mw_state {
- RXE_MW_STATE_INVALID = RXE_MR_STATE_INVALID,
- RXE_MW_STATE_FREE = RXE_MR_STATE_FREE,
- RXE_MW_STATE_VALID = RXE_MR_STATE_VALID,
+ RXE_MW_STATE_INVALID = RXE_MR_STATE_INVALID,
+ RXE_MW_STATE_FREE = RXE_MR_STATE_FREE,
+ RXE_MW_STATE_VALID = RXE_MR_STATE_VALID,
};
struct rxe_mw {
- struct ib_mw ibmw;
- struct rxe_pool_entry pelem;
+ struct ib_mw ibmw;
+ struct rxe_pool_entry pelem;
spinlock_t lock;
enum rxe_mw_state state;
- struct rxe_qp *qp; /* Type 2 only */
+ struct rxe_qp *qp; /* Type 2 only */
struct rxe_mr *mr;
int access;
u64 addr;
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH for-next 9/9] RDMA/rxe: Implement memory access through MWs
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (7 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 8/9] RDMA/rxe: Implement invalidate MW operations Bob Pearson
@ 2021-04-08 21:40 ` Bob Pearson
2021-04-13 23:22 ` [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Jason Gunthorpe
9 siblings, 0 replies; 13+ messages in thread
From: Bob Pearson @ 2021-04-08 21:40 UTC (permalink / raw)
To: jgg, xyzxyz2000, linux-rdma; +Cc: Bob Pearson
Add code to implement memory access through memory windows.
Signed-off-by: Bob Pearson <rpearson@hpe.com>
---
drivers/infiniband/sw/rxe/rxe_loc.h | 1 +
drivers/infiniband/sw/rxe/rxe_mw.c | 23 +++++++++++
drivers/infiniband/sw/rxe/rxe_resp.c | 58 +++++++++++++++++++--------
drivers/infiniband/sw/rxe/rxe_verbs.h | 11 +++++
4 files changed, 77 insertions(+), 16 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_loc.h b/drivers/infiniband/sw/rxe/rxe_loc.h
index 7f1117c51e30..99158d11dae7 100644
--- a/drivers/infiniband/sw/rxe/rxe_loc.h
+++ b/drivers/infiniband/sw/rxe/rxe_loc.h
@@ -104,6 +104,7 @@ int rxe_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata);
int rxe_dealloc_mw(struct ib_mw *ibmw);
int rxe_bind_mw(struct rxe_qp *qp, struct rxe_send_wqe *wqe);
int rxe_invalidate_mw(struct rxe_qp *qp, u32 rkey);
+struct rxe_mw *lookup_mw(struct rxe_qp *qp, int access, u32 rkey);
void rxe_mw_cleanup(struct rxe_pool_entry *arg);
/* rxe_net.c */
diff --git a/drivers/infiniband/sw/rxe/rxe_mw.c b/drivers/infiniband/sw/rxe/rxe_mw.c
index 7b39670f32cd..5c26de154c6e 100644
--- a/drivers/infiniband/sw/rxe/rxe_mw.c
+++ b/drivers/infiniband/sw/rxe/rxe_mw.c
@@ -325,6 +325,29 @@ int rxe_invalidate_mw(struct rxe_qp *qp, u32 rkey)
return ret;
}
+struct rxe_mw *lookup_mw(struct rxe_qp *qp, int access, u32 rkey)
+{
+ struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
+ struct rxe_pd *pd = to_rpd(qp->ibqp.pd);
+ struct rxe_mw *mw;
+ int index = rkey >> 8;
+
+ mw = rxe_pool_get_index(&rxe->mw_pool, index);
+ if (!mw)
+ return NULL;
+
+ if (unlikely((mw_rkey(mw) != rkey) || mw_pd(mw) != pd ||
+ (mw->ibmw.type == IB_MW_TYPE_2 && mw->qp != qp) ||
+ (mw->length == 0) ||
+ (access && !(access & mw->access)) ||
+ mw->state != RXE_MW_STATE_VALID)) {
+ rxe_drop_ref(mw);
+ return NULL;
+ }
+
+ return mw;
+}
+
void rxe_mw_cleanup(struct rxe_pool_entry *elem)
{
struct rxe_mw *mw = container_of(elem, typeof(*mw), pelem);
diff --git a/drivers/infiniband/sw/rxe/rxe_resp.c b/drivers/infiniband/sw/rxe/rxe_resp.c
index 8834f8a23644..dd0b4cb9dde0 100644
--- a/drivers/infiniband/sw/rxe/rxe_resp.c
+++ b/drivers/infiniband/sw/rxe/rxe_resp.c
@@ -394,6 +394,7 @@ static enum resp_states check_rkey(struct rxe_qp *qp,
struct rxe_pkt_info *pkt)
{
struct rxe_mr *mr = NULL;
+ struct rxe_mw *mw = NULL;
u64 va;
u32 rkey;
u32 resid;
@@ -405,6 +406,7 @@ static enum resp_states check_rkey(struct rxe_qp *qp,
if (pkt->mask & (RXE_READ_MASK | RXE_WRITE_MASK)) {
if (pkt->mask & RXE_RETH_MASK) {
qp->resp.va = reth_va(pkt);
+ qp->resp.offset = 0;
qp->resp.rkey = reth_rkey(pkt);
qp->resp.resid = reth_len(pkt);
qp->resp.length = reth_len(pkt);
@@ -413,6 +415,7 @@ static enum resp_states check_rkey(struct rxe_qp *qp,
: IB_ACCESS_REMOTE_WRITE;
} else if (pkt->mask & RXE_ATOMIC_MASK) {
qp->resp.va = atmeth_va(pkt);
+ qp->resp.offset = 0;
qp->resp.rkey = atmeth_rkey(pkt);
qp->resp.resid = sizeof(u64);
access = IB_ACCESS_REMOTE_ATOMIC;
@@ -432,18 +435,36 @@ static enum resp_states check_rkey(struct rxe_qp *qp,
resid = qp->resp.resid;
pktlen = payload_size(pkt);
- mr = lookup_mr(qp->pd, access, rkey, lookup_remote);
- if (!mr) {
- state = RESPST_ERR_RKEY_VIOLATION;
- goto err;
- }
+ if (rkey_is_mw(rkey)) {
+ mw = lookup_mw(qp, access, rkey);
+ if (!mw) {
+ pr_err("%s: no MW matches rkey %#x\n", __func__, rkey);
+ state = RESPST_ERR_RKEY_VIOLATION;
+ goto err;
+ }
- if (unlikely(mr->state == RXE_MR_STATE_FREE)) {
- state = RESPST_ERR_RKEY_VIOLATION;
- goto err;
+ mr = mw->mr;
+ if (!mr) {
+ pr_err("%s: MW doesn't have an MR\n", __func__);
+ state = RESPST_ERR_RKEY_VIOLATION;
+ goto err;
+ }
+
+ if (mw->access & IB_ZERO_BASED)
+ qp->resp.offset = mw->addr;
+
+ rxe_drop_ref(mw);
+ rxe_add_ref(mr);
+ } else {
+ mr = lookup_mr(qp->pd, access, rkey, lookup_remote);
+ if (!mr) {
+ pr_err("%s: no MR matches rkey %#x\n", __func__, rkey);
+ state = RESPST_ERR_RKEY_VIOLATION;
+ goto err;
+ }
}
- if (mr_check_range(mr, va, resid)) {
+ if (mr_check_range(mr, va + qp->resp.offset, resid)) {
state = RESPST_ERR_RKEY_VIOLATION;
goto err;
}
@@ -477,6 +498,9 @@ static enum resp_states check_rkey(struct rxe_qp *qp,
err:
if (mr)
rxe_drop_ref(mr);
+ if (mw)
+ rxe_drop_ref(mw);
+
return state;
}
@@ -501,8 +525,8 @@ static enum resp_states write_data_in(struct rxe_qp *qp,
int err;
int data_len = payload_size(pkt);
- err = rxe_mr_copy(qp->resp.mr, qp->resp.va, payload_addr(pkt), data_len,
- to_mr_obj, NULL);
+ err = rxe_mr_copy(qp->resp.mr, qp->resp.va + qp->resp.offset,
+ payload_addr(pkt), data_len, to_mr_obj, NULL);
if (err) {
rc = RESPST_ERR_RKEY_VIOLATION;
goto out;
@@ -521,7 +545,6 @@ static DEFINE_SPINLOCK(atomic_ops_lock);
static enum resp_states process_atomic(struct rxe_qp *qp,
struct rxe_pkt_info *pkt)
{
- u64 iova = atmeth_va(pkt);
u64 *vaddr;
enum resp_states ret;
struct rxe_mr *mr = qp->resp.mr;
@@ -531,7 +554,7 @@ static enum resp_states process_atomic(struct rxe_qp *qp,
goto out;
}
- vaddr = iova_to_vaddr(mr, iova, sizeof(u64));
+ vaddr = iova_to_vaddr(mr, qp->resp.va + qp->resp.offset, sizeof(u64));
/* check vaddr is 8 bytes aligned. */
if (!vaddr || (uintptr_t)vaddr & 7) {
@@ -655,8 +678,10 @@ static enum resp_states read_reply(struct rxe_qp *qp,
res->type = RXE_READ_MASK;
res->replay = 0;
- res->read.va = qp->resp.va;
- res->read.va_org = qp->resp.va;
+ res->read.va = qp->resp.va +
+ qp->resp.offset;
+ res->read.va_org = qp->resp.va +
+ qp->resp.offset;
res->first_psn = req_pkt->psn;
@@ -835,7 +860,7 @@ static enum resp_states do_complete(struct rxe_qp *qp,
struct rxe_dev *rxe = to_rdev(qp->ibqp.device);
if (unlikely(!wqe))
- return RESPST_CLEANUP;
+ goto done;
memset(&cqe, 0, sizeof(cqe));
@@ -925,6 +950,7 @@ static enum resp_states do_complete(struct rxe_qp *qp,
if (qp->resp.state == QP_STATE_ERROR)
return RESPST_CHK_RESOURCE;
+done:
if (!pkt)
return RESPST_DONE;
else if (qp_type(qp) == IB_QPT_RC)
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h
index b286a14ec282..9f35e2c042d0 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.h
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.h
@@ -183,6 +183,7 @@ struct rxe_resp_info {
/* RDMA read / atomic only */
u64 va;
+ u64 offset;
struct rxe_mr *mr;
u32 resid;
u32 rkey;
@@ -470,6 +471,16 @@ static inline u32 mr_rkey(struct rxe_mr *mr)
return mr->ibmr.rkey;
}
+static inline struct rxe_pd *mw_pd(struct rxe_mw *mw)
+{
+ return to_rpd(mw->ibmw.pd);
+}
+
+static inline u32 mw_rkey(struct rxe_mw *mw)
+{
+ return mw->ibmw.rkey;
+}
+
int rxe_register_device(struct rxe_dev *rxe, const char *ibdev_name);
void rxe_mc_cleanup(struct rxe_pool_entry *arg);
--
2.27.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH for-next 0/9] RDMA/rxe: Implement memory windows
2021-04-08 21:40 [PATCH for-next 0/9] RDMA/rxe: Implement memory windows Bob Pearson
` (8 preceding siblings ...)
2021-04-08 21:40 ` [PATCH for-next 9/9] RDMA/rxe: Implement memory access through MWs Bob Pearson
@ 2021-04-13 23:22 ` Jason Gunthorpe
9 siblings, 0 replies; 13+ messages in thread
From: Jason Gunthorpe @ 2021-04-13 23:22 UTC (permalink / raw)
To: Bob Pearson; +Cc: xyzxyz2000, linux-rdma, Bob Pearson
On Thu, Apr 08, 2021 at 04:40:32PM -0500, Bob Pearson wrote:
> This series of patches implement memory windows for the rdma_rxe
> driver. This is a shorter reimplementation of an earlier patch
> set. They apply to and depend on the current for-next linux rdma
> tree.
>
> Bob Pearson (9):
> RDMA/rxe: Add bind MW fields to rxe_send_wr
> RDMA/rxe: Return errors for add index and key
> RDMA/rxe: Enable MW object pool
> RDMA/rxe: Add ib_alloc_mw and ib_dealloc_mw verbs
> RDMA/rxe: Replace WR_REG_MASK by WR_LOCAL_OP_MASK
> RDMA/rxe: Move local ops to subroutine
> RDMA/rxe: Add support for bind MW work requests
> RDMA/rxe: Implement invalidate MW operations
> RDMA/rxe: Implement memory access through MWs
>
> drivers/infiniband/sw/rxe/Makefile | 1 +
> drivers/infiniband/sw/rxe/rxe.c | 1 +
> drivers/infiniband/sw/rxe/rxe_comp.c | 1 +
> drivers/infiniband/sw/rxe/rxe_loc.h | 29 +-
> drivers/infiniband/sw/rxe/rxe_mr.c | 79 ++++--
> drivers/infiniband/sw/rxe/rxe_mw.c | 356 +++++++++++++++++++++++++
> drivers/infiniband/sw/rxe/rxe_opcode.c | 11 +-
> drivers/infiniband/sw/rxe/rxe_opcode.h | 3 +-
> drivers/infiniband/sw/rxe/rxe_param.h | 19 +-
> drivers/infiniband/sw/rxe/rxe_pool.c | 45 ++--
> drivers/infiniband/sw/rxe/rxe_pool.h | 8 +-
> drivers/infiniband/sw/rxe/rxe_req.c | 102 ++++---
> drivers/infiniband/sw/rxe/rxe_resp.c | 110 +++++---
> drivers/infiniband/sw/rxe/rxe_verbs.c | 5 +-
> drivers/infiniband/sw/rxe/rxe_verbs.h | 38 ++-
> include/uapi/rdma/rdma_user_rxe.h | 34 ++-
> 16 files changed, 691 insertions(+), 151 deletions(-)
> create mode 100644 drivers/infiniband/sw/rxe/rxe_mw.c
I think the uapi file could be tidied enough with some #ifdef
__KERNEL__, but the series doesn't apply anyhow:
Auto-merging drivers/infiniband/sw/rxe/rxe_resp.c
CONFLICT (content): Merge conflict in drivers/infiniband/sw/rxe/rxe_resp.c
Jason
^ permalink raw reply [flat|nested] 13+ messages in thread