| Message ID | 1548768386-28289-6-git-send-email-joeln@il.ibm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | RDMA: reg_remote_mr | expand |
On Tue, Jan 29, 2019 at 03:26:26PM +0200, Joel Nider wrote: > Add a new handler for new uverb reg_remote_mr. The purpose is to register > a memory region in a different address space (i.e. process) than the > caller. > > The main use case which motivated this change is post-copy container > migration. When a migration manager (i.e. CRIU) starts a migration, it > must have an open connection for handling any page faults that occur > in the container after restoration on the target machine. Even though > CRIU establishes and maintains the connection, ultimately the memory > is copied from the container being migrated (i.e. a remote address > space). This container must remain passive -- meaning it cannot have > any knowledge of the RDMA connection; therefore the migration manager > must have the ability to register a remote memory region. This remote > memory region will serve as the source for any memory pages that must > be copied (on-demand or otherwise) during the migration. > > Signed-off-by: Joel Nider <joeln@il.ibm.com> > drivers/infiniband/core/uverbs_std_types_mr.c | 129 +++++++++++++++++++++++++- > include/rdma/ib_verbs.h | 8 ++ > include/uapi/rdma/ib_user_ioctl_cmds.h | 13 +++ > 3 files changed, 149 insertions(+), 1 deletion(-) > > diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/infiniband/core/uverbs_std_types_mr.c > index 4d4be0c..bf7b4b2 100644 > +++ b/drivers/infiniband/core/uverbs_std_types_mr.c > @@ -150,6 +150,99 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( > return ret; > } > > +static int UVERBS_HANDLER(UVERBS_METHOD_REG_REMOTE_MR)( > + struct uverbs_attr_bundle *attrs) > +{ I think this should just be REG_MR with an optional remote PID argument > + struct pid *owner_pid; > + struct ib_reg_remote_mr_attr attr = {}; > + struct ib_uobject *uobj = > + uverbs_attr_get_uobject(attrs, > + UVERBS_ATTR_REG_REMOTE_MR_HANDLE); > + struct ib_pd *pd = > + uverbs_attr_get_obj(attrs, UVERBS_ATTR_REG_REMOTE_MR_PD_HANDLE); > + > + struct ib_mr *mr; > + int ret; > + > + ret = uverbs_copy_from(&attr.start, attrs, > + UVERBS_ATTR_REG_REMOTE_MR_START); > + if (ret) > + return ret; > + > + ret = uverbs_copy_from(&attr.length, attrs, > + UVERBS_ATTR_REG_REMOTE_MR_LENGTH); > + if (ret) > + return ret; > + > + ret = uverbs_copy_from(&attr.hca_va, attrs, > + UVERBS_ATTR_REG_REMOTE_MR_HCA_VA); > + if (ret) > + return ret; > + > + ret = uverbs_copy_from(&attr.owner, attrs, > + UVERBS_ATTR_REG_REMOTE_MR_OWNER); > + if (ret) > + return ret; Maybe these should use the const version, it is becoming intended for small integers, then we can do sensible things like use uintptr_t to store pointer values, and size_t to store sizes - the code will automatically bounds check the user input if it is done like this. > + ret = uverbs_get_flags32(&attr.access_flags, attrs, > + UVERBS_ATTR_REG_REMOTE_MR_ACCESS_FLAGS, > + IB_ACCESS_SUPPORTED); > + if (ret) > + return ret; > + > + /* ensure the offsets are identical */ > + if ((attr.start & ~PAGE_MASK) != (attr.hca_va & ~PAGE_MASK)) > + return -EINVAL; > + > + ret = ib_check_mr_access(attr.access_flags); > + if (ret) > + return ret; > + > + if (attr.access_flags & IB_ACCESS_ON_DEMAND) { > + if (!(pd->device->attrs.device_cap_flags & > + IB_DEVICE_ON_DEMAND_PAGING)) { > + pr_debug("ODP support not available\n"); > + ret = -EINVAL; > + return ret; > + } > + } > + > + /* get the owner's pid struct before something happens to it */ > + owner_pid = find_get_pid(attr.owner); security? Match what ptrace does? > + mr = pd->device->ops.reg_user_mr(pd, attr.start, attr.length, > + attr.hca_va, attr.access_flags, owner_pid, NULL); > + if (IS_ERR(mr)) > + return PTR_ERR(mr); > + > + mr->device = pd->device; > + mr->pd = pd; > + mr->dm = NULL; > + mr->uobject = uobj; > + atomic_inc(&pd->usecnt); > + mr->res.type = RDMA_RESTRACK_MR; > + mr->res.task = get_pid_task(owner_pid, PIDTYPE_PID); > + rdma_restrack_kadd(&mr->res); > + > + uobj->object = mr; > + > + ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_REMOTE_MR_RESP_LKEY, > + &mr->lkey, sizeof(mr->lkey)); > + if (ret) > + goto err_dereg; > + > + ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_REMOTE_MR_RESP_RKEY, > + &mr->rkey, sizeof(mr->rkey)); > + if (ret) > + goto err_dereg; > + > + return 0; > + > +err_dereg: > + ib_dereg_mr(mr); > + > + return ret; > +} > + > DECLARE_UVERBS_NAMED_METHOD( > UVERBS_METHOD_ADVISE_MR, > UVERBS_ATTR_IDR(UVERBS_ATTR_ADVISE_MR_PD_HANDLE, > @@ -203,12 +296,46 @@ DECLARE_UVERBS_NAMED_METHOD_DESTROY( > UVERBS_ACCESS_DESTROY, > UA_MANDATORY)); > > +DECLARE_UVERBS_NAMED_METHOD( > + UVERBS_METHOD_REG_REMOTE_MR, > + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_REMOTE_MR_HANDLE, > + UVERBS_OBJECT_MR, > + UVERBS_ACCESS_NEW, > + UA_MANDATORY), > + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_REMOTE_MR_PD_HANDLE, > + UVERBS_OBJECT_PD, > + UVERBS_ACCESS_READ, > + UA_MANDATORY), > + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_START, > + UVERBS_ATTR_TYPE(u64), > + UA_MANDATORY), > + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_LENGTH, > + UVERBS_ATTR_TYPE(u64), > + UA_MANDATORY), > + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_HCA_VA, > + UVERBS_ATTR_TYPE(u64), > + UA_MANDATORY), > + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_REG_REMOTE_MR_ACCESS_FLAGS, > + enum ib_access_flags), > + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_OWNER, > + UVERBS_ATTR_TYPE(u32), > + UA_MANDATORY), > + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_REMOTE_MR_RESP_LKEY, > + UVERBS_ATTR_TYPE(u32), > + UA_MANDATORY), > + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_REMOTE_MR_RESP_RKEY, > + UVERBS_ATTR_TYPE(u32), > + UA_MANDATORY), > +); > + > DECLARE_UVERBS_NAMED_OBJECT( > UVERBS_OBJECT_MR, > UVERBS_TYPE_ALLOC_IDR(uverbs_free_mr), > &UVERBS_METHOD(UVERBS_METHOD_DM_MR_REG), > &UVERBS_METHOD(UVERBS_METHOD_MR_DESTROY), > - &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR)); > + &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR), > + &UVERBS_METHOD(UVERBS_METHOD_REG_REMOTE_MR), > +); I'm kind of surprised this compiles with the trailing comma? > const struct uapi_definition uverbs_def_obj_mr[] = { > UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_MR, > diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h > index 3432404..dcf5edc 100644 > +++ b/include/rdma/ib_verbs.h > @@ -334,6 +334,14 @@ struct ib_dm_alloc_attr { > u32 flags; > }; > > +struct ib_reg_remote_mr_attr { > + u64 start; > + u64 length; > + u64 hca_va; > + u32 access_flags; > + u32 owner; > +}; Why? Why here? Jason
linux-rdma-owner@vger.kernel.org wrote on 01/29/2019 07:04:06 PM: > On Tue, Jan 29, 2019 at 03:26:26PM +0200, Joel Nider wrote: > > Add a new handler for new uverb reg_remote_mr. The purpose is to register > > a memory region in a different address space (i.e. process) than the > > caller. > > > > The main use case which motivated this change is post-copy container > > migration. When a migration manager (i.e. CRIU) starts a migration, it > > must have an open connection for handling any page faults that occur > > in the container after restoration on the target machine. Even though > > CRIU establishes and maintains the connection, ultimately the memory > > is copied from the container being migrated (i.e. a remote address > > space). This container must remain passive -- meaning it cannot have > > any knowledge of the RDMA connection; therefore the migration manager > > must have the ability to register a remote memory region. This remote > > memory region will serve as the source for any memory pages that must > > be copied (on-demand or otherwise) during the migration. > > > > Signed-off-by: Joel Nider <joeln@il.ibm.com> > > drivers/infiniband/core/uverbs_std_types_mr.c | 129 +++++++++++++++++++++++++- > > include/rdma/ib_verbs.h | 8 ++ > > include/uapi/rdma/ib_user_ioctl_cmds.h | 13 +++ > > 3 files changed, 149 insertions(+), 1 deletion(-) > > > > diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/ > infiniband/core/uverbs_std_types_mr.c > > index 4d4be0c..bf7b4b2 100644 > > +++ b/drivers/infiniband/core/uverbs_std_types_mr.c > > @@ -150,6 +150,99 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( > > return ret; > > } > > > > +static int UVERBS_HANDLER(UVERBS_METHOD_REG_REMOTE_MR)( > > + struct uverbs_attr_bundle *attrs) > > +{ > > I think this should just be REG_MR with an optional remote PID > argument Maybe I missed something. Isn't REG_MR only implemented as a write() command? In our earlier conversation you told me all new commands must be implemented as ioctl() commands. > > DECLARE_UVERBS_NAMED_OBJECT( > > UVERBS_OBJECT_MR, > > UVERBS_TYPE_ALLOC_IDR(uverbs_free_mr), > > &UVERBS_METHOD(UVERBS_METHOD_DM_MR_REG), > > &UVERBS_METHOD(UVERBS_METHOD_MR_DESTROY), > > - &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR)); > > + &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR), > > + &UVERBS_METHOD(UVERBS_METHOD_REG_REMOTE_MR), > > +); > > I'm kind of surprised this compiles with the trailing comma? Personally, I think it is nicer with the trailing comma. Of course syntactically it makes no sense, but when adding a new entry, you don't have to touch the previous line, which makes the diff cleaner. If this is against standard practices I will remove the comma.
On Wed, Jan 30, 2019 at 10:34:02AM +0200, Joel Nider wrote: > linux-rdma-owner@vger.kernel.org wrote on 01/29/2019 07:04:06 PM: > > > On Tue, Jan 29, 2019 at 03:26:26PM +0200, Joel Nider wrote: > > > Add a new handler for new uverb reg_remote_mr. The purpose is to > register > > > a memory region in a different address space (i.e. process) than the > > > caller. > > > > > > The main use case which motivated this change is post-copy container > > > migration. When a migration manager (i.e. CRIU) starts a migration, it > > > must have an open connection for handling any page faults that occur > > > in the container after restoration on the target machine. Even though > > > CRIU establishes and maintains the connection, ultimately the memory > > > is copied from the container being migrated (i.e. a remote address > > > space). This container must remain passive -- meaning it cannot have > > > any knowledge of the RDMA connection; therefore the migration manager > > > must have the ability to register a remote memory region. This remote > > > memory region will serve as the source for any memory pages that must > > > be copied (on-demand or otherwise) during the migration. > > > > > > Signed-off-by: Joel Nider <joeln@il.ibm.com> > > > drivers/infiniband/core/uverbs_std_types_mr.c | 129 > +++++++++++++++++++++++++- > > > include/rdma/ib_verbs.h | 8 ++ > > > include/uapi/rdma/ib_user_ioctl_cmds.h | 13 +++ > > > 3 files changed, 149 insertions(+), 1 deletion(-) > > > > > > diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/ > > infiniband/core/uverbs_std_types_mr.c > > > index 4d4be0c..bf7b4b2 100644 > > > +++ b/drivers/infiniband/core/uverbs_std_types_mr.c > > > @@ -150,6 +150,99 @@ static int > UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( > > > return ret; > > > } > > > > > > +static int UVERBS_HANDLER(UVERBS_METHOD_REG_REMOTE_MR)( > > > + struct uverbs_attr_bundle *attrs) > > > +{ > > > > I think this should just be REG_MR with an optional remote PID > > argument > > Maybe I missed something. Isn't REG_MR only implemented as a write() > command? In our earlier conversation you told me all new commands must be > implemented as ioctl() commands. Yes - but we are also converting old write() commands into ioctl() when they need new functionality. So in this case it should convert reg_mr to ioctl() then add an optional report PID argument > > > > DECLARE_UVERBS_NAMED_OBJECT( > > > UVERBS_OBJECT_MR, > > > UVERBS_TYPE_ALLOC_IDR(uverbs_free_mr), > > > &UVERBS_METHOD(UVERBS_METHOD_DM_MR_REG), > > > &UVERBS_METHOD(UVERBS_METHOD_MR_DESTROY), > > > - &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR)); > > > + &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR), > > > + &UVERBS_METHOD(UVERBS_METHOD_REG_REMOTE_MR), > > > +); > > > > I'm kind of surprised this compiles with the trailing comma? > Personally, I think it is nicer with the trailing comma. Of course > syntactically it makes no sense, but when adding a new entry, you don't > have to touch the previous line, which makes the diff cleaner. If this is > against standard practices I will remove the comma. Well, it is just that this is a macro call, and you usually can't have a trailing comma in a function-macro call, at least I thought this was the case.. Without some study I'm not sure what it expands to, or if that expansion is even OK.. Jason
diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/infiniband/core/uverbs_std_types_mr.c index 4d4be0c..bf7b4b2 100644 --- a/drivers/infiniband/core/uverbs_std_types_mr.c +++ b/drivers/infiniband/core/uverbs_std_types_mr.c @@ -150,6 +150,99 @@ static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( return ret; } +static int UVERBS_HANDLER(UVERBS_METHOD_REG_REMOTE_MR)( + struct uverbs_attr_bundle *attrs) +{ + struct pid *owner_pid; + struct ib_reg_remote_mr_attr attr = {}; + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, + UVERBS_ATTR_REG_REMOTE_MR_HANDLE); + struct ib_pd *pd = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_REG_REMOTE_MR_PD_HANDLE); + + struct ib_mr *mr; + int ret; + + ret = uverbs_copy_from(&attr.start, attrs, + UVERBS_ATTR_REG_REMOTE_MR_START); + if (ret) + return ret; + + ret = uverbs_copy_from(&attr.length, attrs, + UVERBS_ATTR_REG_REMOTE_MR_LENGTH); + if (ret) + return ret; + + ret = uverbs_copy_from(&attr.hca_va, attrs, + UVERBS_ATTR_REG_REMOTE_MR_HCA_VA); + if (ret) + return ret; + + ret = uverbs_copy_from(&attr.owner, attrs, + UVERBS_ATTR_REG_REMOTE_MR_OWNER); + if (ret) + return ret; + + ret = uverbs_get_flags32(&attr.access_flags, attrs, + UVERBS_ATTR_REG_REMOTE_MR_ACCESS_FLAGS, + IB_ACCESS_SUPPORTED); + if (ret) + return ret; + + /* ensure the offsets are identical */ + if ((attr.start & ~PAGE_MASK) != (attr.hca_va & ~PAGE_MASK)) + return -EINVAL; + + ret = ib_check_mr_access(attr.access_flags); + if (ret) + return ret; + + if (attr.access_flags & IB_ACCESS_ON_DEMAND) { + if (!(pd->device->attrs.device_cap_flags & + IB_DEVICE_ON_DEMAND_PAGING)) { + pr_debug("ODP support not available\n"); + ret = -EINVAL; + return ret; + } + } + + /* get the owner's pid struct before something happens to it */ + owner_pid = find_get_pid(attr.owner); + mr = pd->device->ops.reg_user_mr(pd, attr.start, attr.length, + attr.hca_va, attr.access_flags, owner_pid, NULL); + if (IS_ERR(mr)) + return PTR_ERR(mr); + + mr->device = pd->device; + mr->pd = pd; + mr->dm = NULL; + mr->uobject = uobj; + atomic_inc(&pd->usecnt); + mr->res.type = RDMA_RESTRACK_MR; + mr->res.task = get_pid_task(owner_pid, PIDTYPE_PID); + rdma_restrack_kadd(&mr->res); + + uobj->object = mr; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_REMOTE_MR_RESP_LKEY, + &mr->lkey, sizeof(mr->lkey)); + if (ret) + goto err_dereg; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_REMOTE_MR_RESP_RKEY, + &mr->rkey, sizeof(mr->rkey)); + if (ret) + goto err_dereg; + + return 0; + +err_dereg: + ib_dereg_mr(mr); + + return ret; +} + DECLARE_UVERBS_NAMED_METHOD( UVERBS_METHOD_ADVISE_MR, UVERBS_ATTR_IDR(UVERBS_ATTR_ADVISE_MR_PD_HANDLE, @@ -203,12 +296,46 @@ DECLARE_UVERBS_NAMED_METHOD_DESTROY( UVERBS_ACCESS_DESTROY, UA_MANDATORY)); +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_REG_REMOTE_MR, + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_REMOTE_MR_HANDLE, + UVERBS_OBJECT_MR, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_REMOTE_MR_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_START, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_LENGTH, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_HCA_VA, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_REG_REMOTE_MR_ACCESS_FLAGS, + enum ib_access_flags), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_REMOTE_MR_OWNER, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_REMOTE_MR_RESP_LKEY, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_REMOTE_MR_RESP_RKEY, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), +); + DECLARE_UVERBS_NAMED_OBJECT( UVERBS_OBJECT_MR, UVERBS_TYPE_ALLOC_IDR(uverbs_free_mr), &UVERBS_METHOD(UVERBS_METHOD_DM_MR_REG), &UVERBS_METHOD(UVERBS_METHOD_MR_DESTROY), - &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR)); + &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR), + &UVERBS_METHOD(UVERBS_METHOD_REG_REMOTE_MR), +); const struct uapi_definition uverbs_def_obj_mr[] = { UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_MR, diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 3432404..dcf5edc 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -334,6 +334,14 @@ struct ib_dm_alloc_attr { u32 flags; }; +struct ib_reg_remote_mr_attr { + u64 start; + u64 length; + u64 hca_va; + u32 access_flags; + u32 owner; +}; + struct ib_device_attr { u64 fw_ver; __be64 sys_image_guid; diff --git a/include/uapi/rdma/ib_user_ioctl_cmds.h b/include/uapi/rdma/ib_user_ioctl_cmds.h index 64f0e3a..4e62cd4 100644 --- a/include/uapi/rdma/ib_user_ioctl_cmds.h +++ b/include/uapi/rdma/ib_user_ioctl_cmds.h @@ -150,10 +150,23 @@ enum uverbs_attrs_reg_dm_mr_cmd_attr_ids { UVERBS_ATTR_REG_DM_MR_RESP_RKEY, }; +enum uverbs_attrs_reg_remote_mr_cmd_attr_ids { + UVERBS_ATTR_REG_REMOTE_MR_HANDLE, + UVERBS_ATTR_REG_REMOTE_MR_PD_HANDLE, + UVERBS_ATTR_REG_REMOTE_MR_START, + UVERBS_ATTR_REG_REMOTE_MR_LENGTH, + UVERBS_ATTR_REG_REMOTE_MR_HCA_VA, + UVERBS_ATTR_REG_REMOTE_MR_ACCESS_FLAGS, + UVERBS_ATTR_REG_REMOTE_MR_OWNER, + UVERBS_ATTR_REG_REMOTE_MR_RESP_LKEY, + UVERBS_ATTR_REG_REMOTE_MR_RESP_RKEY, +}; + enum uverbs_methods_mr { UVERBS_METHOD_DM_MR_REG, UVERBS_METHOD_MR_DESTROY, UVERBS_METHOD_ADVISE_MR, + UVERBS_METHOD_REG_REMOTE_MR, }; enum uverbs_attrs_mr_destroy_ids {
Add a new handler for new uverb reg_remote_mr. The purpose is to register a memory region in a different address space (i.e. process) than the caller. The main use case which motivated this change is post-copy container migration. When a migration manager (i.e. CRIU) starts a migration, it must have an open connection for handling any page faults that occur in the container after restoration on the target machine. Even though CRIU establishes and maintains the connection, ultimately the memory is copied from the container being migrated (i.e. a remote address space). This container must remain passive -- meaning it cannot have any knowledge of the RDMA connection; therefore the migration manager must have the ability to register a remote memory region. This remote memory region will serve as the source for any memory pages that must be copied (on-demand or otherwise) during the migration. Signed-off-by: Joel Nider <joeln@il.ibm.com> --- drivers/infiniband/core/uverbs_std_types_mr.c | 129 +++++++++++++++++++++++++- include/rdma/ib_verbs.h | 8 ++ include/uapi/rdma/ib_user_ioctl_cmds.h | 13 +++ 3 files changed, 149 insertions(+), 1 deletion(-)