* [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm
@ 2021-08-25 9:43 Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 1/3] RDMA/hns: Bugfix for data type of dip_idx Wenpeng Liang
` (3 more replies)
0 siblings, 4 replies; 6+ messages in thread
From: Wenpeng Liang @ 2021-08-25 9:43 UTC (permalink / raw)
To: dledford, jgg; +Cc: linux-rdma, linuxarm, liangwenpeng
Fix three dip_idx related errors.
Junxian Huang (3):
RDMA/hns: Bugfix for data type of dip_idx
RDMA/hns: Bugfix for the missing assignment for dip_idx
RDMA/hns: Bugfix for incorrect association between dip_idx and dgid
drivers/infiniband/hw/hns/hns_roce_device.h | 9 ++++++++-
drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 13 +++++++++++--
drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 2 +-
drivers/infiniband/hw/hns/hns_roce_main.c | 8 ++++++--
drivers/infiniband/hw/hns/hns_roce_qp.c | 10 +++++++++-
5 files changed, 35 insertions(+), 7 deletions(-)
--
2.8.1
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH for-next 1/3] RDMA/hns: Bugfix for data type of dip_idx
2021-08-25 9:43 [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Wenpeng Liang
@ 2021-08-25 9:43 ` Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 2/3] RDMA/hns: Bugfix for the missing assignment for dip_idx Wenpeng Liang
` (2 subsequent siblings)
3 siblings, 0 replies; 6+ messages in thread
From: Wenpeng Liang @ 2021-08-25 9:43 UTC (permalink / raw)
To: dledford, jgg; +Cc: linux-rdma, linuxarm, liangwenpeng
From: Junxian Huang <huangjunxian4@hisilicon.com>
dip_idx is associated with qp_num whose data type is u32. However, dip_idx
is incorrectly defined as u8 data in the hns_roce_dip struct, which leads
to data truncation during value assignment.
Fixes: f91696f2f053 ("RDMA/hns: Support congestion control type selection according to the FW")
Signed-off-by: Junxian Huang <huangjunxian4@hisilicon.com>
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
---
drivers/infiniband/hw/hns/hns_roce_hw_v2.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h
index 54c1223..0b91a1a 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h
@@ -1441,7 +1441,7 @@ struct hns_roce_v2_priv {
struct hns_roce_dip {
u8 dgid[GID_LEN_V2];
- u8 dip_idx;
+ u32 dip_idx;
struct list_head node; /* all dips are on a list */
};
--
2.8.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH for-next 2/3] RDMA/hns: Bugfix for the missing assignment for dip_idx
2021-08-25 9:43 [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 1/3] RDMA/hns: Bugfix for data type of dip_idx Wenpeng Liang
@ 2021-08-25 9:43 ` Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 3/3] RDMA/hns: Bugfix for incorrect association between dip_idx and dgid Wenpeng Liang
2021-08-25 17:50 ` [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Jason Gunthorpe
3 siblings, 0 replies; 6+ messages in thread
From: Wenpeng Liang @ 2021-08-25 9:43 UTC (permalink / raw)
To: dledford, jgg; +Cc: linux-rdma, linuxarm, liangwenpeng
From: Junxian Huang <huangjunxian4@hisilicon.com>
When the dgid-dip_idx mapping relationship exists, dip should be assigned.
Fixes: f91696f2f053 ("RDMA/hns: Support congestion control type selection according to the FW")
Signed-off-by: Junxian Huang <huangjunxian4@hisilicon.com>
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
---
drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index d00be78..1ad2a91 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -4511,8 +4511,10 @@ static int get_dip_ctx_idx(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
spin_lock_irqsave(&hr_dev->dip_list_lock, flags);
list_for_each_entry(hr_dip, &hr_dev->dip_list, node) {
- if (!memcmp(grh->dgid.raw, hr_dip->dgid, 16))
+ if (!memcmp(grh->dgid.raw, hr_dip->dgid, 16)) {
+ *dip_idx = hr_dip->dip_idx;
goto out;
+ }
}
/* If no dgid is found, a new dip and a mapping between dgid and
--
2.8.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH for-next 3/3] RDMA/hns: Bugfix for incorrect association between dip_idx and dgid
2021-08-25 9:43 [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 1/3] RDMA/hns: Bugfix for data type of dip_idx Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 2/3] RDMA/hns: Bugfix for the missing assignment for dip_idx Wenpeng Liang
@ 2021-08-25 9:43 ` Wenpeng Liang
2021-08-25 17:50 ` [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Jason Gunthorpe
3 siblings, 0 replies; 6+ messages in thread
From: Wenpeng Liang @ 2021-08-25 9:43 UTC (permalink / raw)
To: dledford, jgg; +Cc: linux-rdma, linuxarm, liangwenpeng
From: Junxian Huang <huangjunxian4@hisilicon.com>
dip_idx and dgid should be a one-to-one mapping relationship, but when
qp_num loops back to the start number, it may happen that two different
dgid are assiociated to the same dip_idx incorrectly.
One solution is to store the qp_num that is not assigned to dip_idx in an
array. When a dip_idx needs to be allocated to a new dgid, an spare qp_num
is extracted and assigned to dip_idx.
Fixes: f91696f2f053 ("RDMA/hns: Support congestion control type selection according to the FW")
Signed-off-by: Junxian Huang <huangjunxian4@hisilicon.com>
Signed-off-by: Yangyang Li <liyangyang20@huawei.com>
Signed-off-by: Wenpeng Liang <liangwenpeng@huawei.com>
---
drivers/infiniband/hw/hns/hns_roce_device.h | 9 ++++++++-
drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 9 ++++++++-
drivers/infiniband/hw/hns/hns_roce_main.c | 8 ++++++--
drivers/infiniband/hw/hns/hns_roce_qp.c | 10 +++++++++-
4 files changed, 31 insertions(+), 5 deletions(-)
diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index 2129da3..9467c39 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -496,6 +496,12 @@ struct hns_roce_bank {
u32 next; /* Next ID to allocate. */
};
+struct hns_roce_idx_table {
+ u32 *spare_idx;
+ u32 head;
+ u32 tail;
+};
+
struct hns_roce_qp_table {
struct hns_roce_hem_table qp_table;
struct hns_roce_hem_table irrl_table;
@@ -504,6 +510,7 @@ struct hns_roce_qp_table {
struct mutex scc_mutex;
struct hns_roce_bank bank[HNS_ROCE_QP_BANK_NUM];
struct mutex bank_mutex;
+ struct hns_roce_idx_table idx_table;
};
struct hns_roce_cq_table {
@@ -1144,7 +1151,7 @@ int hns_roce_mtr_map(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr,
void hns_roce_init_pd_table(struct hns_roce_dev *hr_dev);
void hns_roce_init_mr_table(struct hns_roce_dev *hr_dev);
void hns_roce_init_cq_table(struct hns_roce_dev *hr_dev);
-void hns_roce_init_qp_table(struct hns_roce_dev *hr_dev);
+int hns_roce_init_qp_table(struct hns_roce_dev *hr_dev);
void hns_roce_init_srq_table(struct hns_roce_dev *hr_dev);
void hns_roce_init_xrcd_table(struct hns_roce_dev *hr_dev);
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index 1ad2a91..fcf67db 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -4504,12 +4504,18 @@ static int get_dip_ctx_idx(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
{
const struct ib_global_route *grh = rdma_ah_read_grh(&attr->ah_attr);
struct hns_roce_dev *hr_dev = to_hr_dev(ibqp->device);
+ u32 *spare_idx = hr_dev->qp_table.idx_table.spare_idx;
+ u32 *head = &hr_dev->qp_table.idx_table.head;
+ u32 *tail = &hr_dev->qp_table.idx_table.tail;
struct hns_roce_dip *hr_dip;
unsigned long flags;
int ret = 0;
spin_lock_irqsave(&hr_dev->dip_list_lock, flags);
+ spare_idx[*tail] = ibqp->qp_num;
+ *tail = (*tail == hr_dev->caps.num_qps - 1) ? 0 : (*tail + 1);
+
list_for_each_entry(hr_dip, &hr_dev->dip_list, node) {
if (!memcmp(grh->dgid.raw, hr_dip->dgid, 16)) {
*dip_idx = hr_dip->dip_idx;
@@ -4527,7 +4533,8 @@ static int get_dip_ctx_idx(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
}
memcpy(hr_dip->dgid, grh->dgid.raw, sizeof(grh->dgid.raw));
- hr_dip->dip_idx = *dip_idx = ibqp->qp_num;
+ hr_dip->dip_idx = *dip_idx = spare_idx[*head];
+ *head = (*head == hr_dev->caps.num_qps - 1) ? 0 : (*head + 1);
list_add_tail(&hr_dip->node, &hr_dev->dip_list);
out:
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 6467f8f..b409134 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -746,6 +746,12 @@ static int hns_roce_setup_hca(struct hns_roce_dev *hr_dev)
goto err_uar_table_free;
}
+ ret = hns_roce_init_qp_table(hr_dev);
+ if (ret) {
+ dev_err(dev, "Failed to init qp_table.\n");
+ goto err_uar_table_free;
+ }
+
hns_roce_init_pd_table(hr_dev);
if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_XRC)
@@ -755,8 +761,6 @@ static int hns_roce_setup_hca(struct hns_roce_dev *hr_dev)
hns_roce_init_cq_table(hr_dev);
- hns_roce_init_qp_table(hr_dev);
-
if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_SRQ) {
hns_roce_init_srq_table(hr_dev);
}
diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c
index fd0f71a..0d0d5aa 100644
--- a/drivers/infiniband/hw/hns/hns_roce_qp.c
+++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
@@ -1411,12 +1411,17 @@ bool hns_roce_wq_overflow(struct hns_roce_wq *hr_wq, u32 nreq,
return cur + nreq >= hr_wq->wqe_cnt;
}
-void hns_roce_init_qp_table(struct hns_roce_dev *hr_dev)
+int hns_roce_init_qp_table(struct hns_roce_dev *hr_dev)
{
struct hns_roce_qp_table *qp_table = &hr_dev->qp_table;
unsigned int reserved_from_bot;
unsigned int i;
+ qp_table->idx_table.spare_idx = kcalloc(hr_dev->caps.num_qps,
+ sizeof(u32), GFP_KERNEL);
+ if (!qp_table->idx_table.spare_idx)
+ return -ENOMEM;
+
mutex_init(&qp_table->scc_mutex);
mutex_init(&qp_table->bank_mutex);
xa_init(&hr_dev->qp_table_xa);
@@ -1434,6 +1439,8 @@ void hns_roce_init_qp_table(struct hns_roce_dev *hr_dev)
HNS_ROCE_QP_BANK_NUM - 1;
hr_dev->qp_table.bank[i].next = hr_dev->qp_table.bank[i].min;
}
+
+ return 0;
}
void hns_roce_cleanup_qp_table(struct hns_roce_dev *hr_dev)
@@ -1442,4 +1449,5 @@ void hns_roce_cleanup_qp_table(struct hns_roce_dev *hr_dev)
for (i = 0; i < HNS_ROCE_QP_BANK_NUM; i++)
ida_destroy(&hr_dev->qp_table.bank[i].ida);
+ kfree(hr_dev->qp_table.idx_table.spare_idx);
}
--
2.8.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm
2021-08-25 9:43 [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Wenpeng Liang
` (2 preceding siblings ...)
2021-08-25 9:43 ` [PATCH for-next 3/3] RDMA/hns: Bugfix for incorrect association between dip_idx and dgid Wenpeng Liang
@ 2021-08-25 17:50 ` Jason Gunthorpe
2021-08-26 11:41 ` Wenpeng Liang
3 siblings, 1 reply; 6+ messages in thread
From: Jason Gunthorpe @ 2021-08-25 17:50 UTC (permalink / raw)
To: Wenpeng Liang; +Cc: dledford, linux-rdma, linuxarm
On Wed, Aug 25, 2021 at 05:43:09PM +0800, Wenpeng Liang wrote:
> Fix three dip_idx related errors.
>
> Junxian Huang (3):
> RDMA/hns: Bugfix for data type of dip_idx
> RDMA/hns: Bugfix for the missing assignment for dip_idx
> RDMA/hns: Bugfix for incorrect association between dip_idx and dgid
Applied to for-next
Please do not put a blank line after Fixes:, it needs to be part of
the trailer block to parse properly. I fixed it
Thanks,
Jason
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm
2021-08-25 17:50 ` [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Jason Gunthorpe
@ 2021-08-26 11:41 ` Wenpeng Liang
0 siblings, 0 replies; 6+ messages in thread
From: Wenpeng Liang @ 2021-08-26 11:41 UTC (permalink / raw)
To: Jason Gunthorpe; +Cc: dledford, linux-rdma, linuxarm
On 2021/8/26 1:50, Jason Gunthorpe wrote:
> On Wed, Aug 25, 2021 at 05:43:09PM +0800, Wenpeng Liang wrote:
>> Fix three dip_idx related errors.
>>
>> Junxian Huang (3):
>> RDMA/hns: Bugfix for data type of dip_idx
>> RDMA/hns: Bugfix for the missing assignment for dip_idx
>> RDMA/hns: Bugfix for incorrect association between dip_idx and dgid
>
> Applied to for-next
>
> Please do not put a blank line after Fixes:, it needs to be part of
> the trailer block to parse properly. I fixed it
>
> Thanks,
> Jason
> .
>
Sorry, I forgot to delete the blank line when submitting the commit,
I will pay more attention to it in the future.
Thanks,
Wenpeng
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2021-08-26 11:41 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-25 9:43 [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 1/3] RDMA/hns: Bugfix for data type of dip_idx Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 2/3] RDMA/hns: Bugfix for the missing assignment for dip_idx Wenpeng Liang
2021-08-25 9:43 ` [PATCH for-next 3/3] RDMA/hns: Bugfix for incorrect association between dip_idx and dgid Wenpeng Liang
2021-08-25 17:50 ` [PATCH for-next 0/3] RDMA/hns: Fix some errors in the congestion control algorithm Jason Gunthorpe
2021-08-26 11:41 ` Wenpeng Liang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).