[for-rc,01/11] IB/core: Avoid a calltrace for an unsupported device memory registration

Commit Message

Lijun Ou April 26, 2018, 6:46 a.m. UTC

if device memory(DM) is not supported that the dm flow should be
taken. the calltrace as follows:
[  101.064379] Unable to handle kernel paging request at virtual address 0001bc20
[  101.071589] Mem abort info:
[  101.074370]   ESR = 0x96000006
[  101.077413]   Exception class = DABT (current EL), IL = 32 bits
[  101.083321]   SET = 0, FnV = 0
[  101.086364]   EA = 0, S1PTW = 0
[  101.089494] Data abort info:
[  101.092364]   ISV = 0, ISS = 0x00000006
[  101.096189]   CM = 0, WnR = 0
[  101.099146] user pgtable: 4k pages, 48-bit VAs, pgdp = 00000000cb2914c6
[  101.105749] [000000000001bc20] pgd=00000023ce5ae003, pud=00000023ce058003, pmd=0000000000000000
[  101.114439] Internal error: Oops: 96000006 [#1] PREEMPT SMP
[  101.119998] Modules linked in: hns_roce_hw_v2 hns_roce rdma_ucm rdma_cm iw_cm ib_uverbs ib_cm ib_core
[  101.129215] CPU: 2 PID: 1994 Comm: ib_send_bw Not tainted 4.17.0-rc1-00028-g87ef120-dirty #3
[  101.137638] Hardware name: Huawei D06/D06, BIOS Hisilicon D06 UEFI Nemo 2.0 RC0 - B205 04/19/2018
[  101.146496] pstate: 60400009 (nZCv daif +PAN -UAO)
[  101.151298] pc : ib_dereg_mr+0x12c/0x150 [ib_core]
[  101.156087] lr : ib_dereg_mr+0x124/0x150 [ib_core]
[  101.160863] sp : ffff000015153b70
[  101.164165] x29: ffff000015153b70 x28: 0000000000000000
[  101.169465] x27: 000000000000000d x26: ffff000000b3d000
[  101.174764] x25: ffff8023cbd2ce80 x24: 000000000001bc08
[  101.180063] x23: ffff000000af4a18 x22: ffff000000af4000
[  101.185363] x21: ffff000000af1638 x20: ffff000000af1000
[  101.190662] x19: 0000000000000000 x18: ffff00008924b567
[  101.195961] x17: 0000ffff8c8636d0 x16: ffff000008229968
[  101.201260] x15: 00000000fffffffe x14: 0000000000000004
[  101.206559] x13: ffff00000924b575 x12: ffff00000924d990
[  101.211858] x11: ffff0000090f1000 x10: 0000000005f5e0ff
[  101.217157] x9 : ffff000015153870 x8 : 3631202c726d5f67
[  101.222456] x7 : 657265645f62695b x6 : 0000000000000293
[  101.227756] x5 : 0000000000000000 x4 : 0000000000000000
[  101.233055] x3 : ffffffffffffffff x2 : 000000000001bc20
[  101.238354] x1 : 0000000000000000 x0 : 0000000000000019
[  101.243654] Process ib_send_bw (pid: 1994, stack limit = 0x0000000053eb9878)
[  101.250688] Call trace:
[  101.253134]  ib_dereg_mr+0x12c/0x150 [ib_core]
[  101.257572]  uverbs_free_mr+0x30/0x40 [ib_uverbs]
[  101.262267]  remove_commit_idr_uobject+0x24/0x58 [ib_uverbs]
[  101.267915]  _rdma_remove_commit_uobject+0x2c/0xc0 [ib_uverbs]
[  101.273737]  rdma_remove_commit_uobject+0x38/0x80 [ib_uverbs]
[  101.279473]  ib_uverbs_dereg_mr+0xd4/0x138 [ib_uverbs]
[  101.284601]  ib_uverbs_write+0x3b4/0x408 [ib_uverbs]
[  101.289554]  __vfs_write+0x34/0x140
[  101.293029]  vfs_write+0xa4/0x1b0
[  101.296331]  ksys_write+0x5c/0xc8
[  101.299633]  sys_write+0xc/0x18
[  101.302762]  el0_svc_naked+0x30/0x34
[  101.306324] Code: 5280cc02 95d95b1f 91006302 f9800051 (885f7c40)
[  101.312405] ---[ end trace a9e4d13346a1e97f ]---

Signed-off-by: Lijun Ou <[e-mail]>
Signed-off-by: Leon Romanovsky <[e-mail]>
Tested-by: Lijun Ou <[e-mail]>
---
 drivers/infiniband/core/uverbs_std_types_dm.c | 7 ++++++-
 drivers/infiniband/core/verbs.c               | 3 ---
 2 files changed, 6 insertions(+), 4 deletions(-)

Comments

Leon Romanovsky April 26, 2018, 7:50 a.m. UTC | #1

On Thu, Apr 26, 2018 at 02:46:14PM +0800, Lijun Ou wrote:
>
> Signed-off-by: Lijun Ou <[e-mail]>
> Signed-off-by: Leon Romanovsky <[e-mail]>
> Tested-by: Lijun Ou <[e-mail]>
> ---
>  drivers/infiniband/core/uverbs_std_types_dm.c | 7 ++++++-
>  drivers/infiniband/core/verbs.c               | 3 ---
>  2 files changed, 6 insertions(+), 4 deletions(-)
>

Lijun, sorry for not updating you, but this patch is incorrect.

Ariel is working to provide proper fix.

Thanks

Lijun Ou April 26, 2018, 8:08 a.m. UTC | #2

在 2018/4/26 15:50, Leon Romanovsky 写道:
> On Thu, Apr 26, 2018 at 02:46:14PM +0800, Lijun Ou wrote:
>>
>> Signed-off-by: Lijun Ou <[e-mail]>
>> Signed-off-by: Leon Romanovsky <[e-mail]>
>> Tested-by: Lijun Ou <[e-mail]>
>> ---
>>  drivers/infiniband/core/uverbs_std_types_dm.c | 7 ++++++-
>>  drivers/infiniband/core/verbs.c               | 3 ---
>>  2 files changed, 6 insertions(+), 4 deletions(-)
>>
> 
> Lijun, sorry for not updating you, but this patch is incorrect.
> 
> Ariel is working to provide proper fix.
> 
> Thanks
> 
Ok. I am tested gived by your modification. if Ariel will do, please ignore it.

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/infiniband/core/uverbs_std_types_dm.c b/drivers/infiniband/core/uverbs_std_types_dm.c
index 8b68157..b83ccbd 100644
--- a/drivers/infiniband/core/uverbs_std_types_dm.c
+++ b/drivers/infiniband/core/uverbs_std_types_dm.c
@@ -37,11 +37,16 @@  static int uverbs_free_dm(struct ib_uobject *uobject,
 			  enum rdma_remove_reason why)
 {
 	struct ib_dm *dm = uobject->object;
+	int ret;
 
 	if (why == RDMA_REMOVE_DESTROY && atomic_read(&dm->usecnt))
 		return -EBUSY;
 
-	return dm->device->dealloc_dm(dm);
+	ret = dm->device->dealloc_dm(dm);
+	if (!ret)
+		atomic_dec(&dm->usecnt);
+
+	return ret;
 }
 
 static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)(struct ib_device *ib_dev,
diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index 7eff3ae..50a3b20 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -1616,15 +1616,12 @@  int ib_resize_cq(struct ib_cq *cq, int cqe)
 int ib_dereg_mr(struct ib_mr *mr)
 {
 	struct ib_pd *pd = mr->pd;
-	struct ib_dm *dm = mr->dm;
 	int ret;
 
 	rdma_restrack_del(&mr->res);
 	ret = mr->device->dereg_mr(mr);
 	if (!ret) {
 		atomic_dec(&pd->usecnt);
-		if (dm)
-			atomic_dec(&dm->usecnt);
 	}
 
 	return ret;