[PATCH AUTOSEL 5.5 072/121] RDMA/ucma: Put a lock around every call to the rdma_cm layer

From: Sasha Levin
Date: Sat Apr 11 2020 - 19:36:55 EST

Next message: Sasha Levin: "[PATCH AUTOSEL 5.5 071/121] drm/omap: dss: Cleanup DSS ports on initialisation failure"
Previous message: Sasha Levin: "[PATCH AUTOSEL 5.5 062/121] drm/amdgpu: fix parentheses in amdgpu_vm_update_ptes"
In reply to: Sasha Levin: "[PATCH AUTOSEL 5.5 062/121] drm/amdgpu: fix parentheses in amdgpu_vm_update_ptes"
Next in thread: Sasha Levin: "[PATCH AUTOSEL 5.5 071/121] drm/omap: dss: Cleanup DSS ports on initialisation failure"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

From: Jason Gunthorpe <jgg@xxxxxxxxxxxx>

[ Upstream commit 7c11910783a1ea17e88777552ef146cace607b3c ]

The rdma_cm must be used single threaded.

This appears to be a bug in the design, as it does have lots of locking
that seems like it should allow concurrency. However, when it is all said
and done every single place that uses the cma_exch() scheme is broken, and
all the unlocked reads from the ucma of the cm_id data are wrong too.

syzkaller has been finding endless bugs related to this.

Fixing this in any elegant way is some enormous amount of work. Take a
very big hammer and put a mutex around everything to do with the
ucma_context at the top of every syscall.

Fixes: 75216638572f ("RDMA/cma: Export rdma cm interface to userspace")
Link: https://lore.kernel.org/r/20200218210432.GA31966@xxxxxxxx
Reported-by: syzbot+adb15cf8c2798e4e0db4@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+e5579222b6a3edd96522@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+4b628fcc748474003457@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+29ee8f76017ce6cf03da@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+6956235342b7317ec564@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+b358909d8d01556b790b@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+6b46b135602a3f3ac99e@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+8458d13b13562abf6b77@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+bd034f3fdc0402e942ed@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+c92378b32760a4eef756@xxxxxxxxxxxxxxxxxxxxxxxxx
Reported-by: syzbot+68b44a1597636e0b342c@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Jason Gunthorpe <jgg@xxxxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
drivers/infiniband/core/ucma.c | 49 ++++++++++++++++++++++++++++++++--
1 file changed, 47 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c
index 0274e9b704be5..f4f79f1292b91 100644
--- a/drivers/infiniband/core/ucma.c
+++ b/drivers/infiniband/core/ucma.c
@@ -91,6 +91,7 @@ struct ucma_context {

struct ucma_file *file;
struct rdma_cm_id *cm_id;
+ struct mutex mutex;
u64 uid;

struct list_head list;
@@ -216,6 +217,7 @@ static struct ucma_context *ucma_alloc_ctx(struct ucma_file *file)
init_completion(&ctx->comp);
INIT_LIST_HEAD(&ctx->mc_list);
ctx->file = file;
+ mutex_init(&ctx->mutex);

if (xa_alloc(&ctx_table, &ctx->id, ctx, xa_limit_32b, GFP_KERNEL))
goto error;
@@ -589,6 +591,7 @@ static int ucma_free_ctx(struct ucma_context *ctx)
}

events_reported = ctx->events_reported;
+ mutex_destroy(&ctx->mutex);
kfree(ctx);
return events_reported;
}
@@ -658,7 +661,10 @@ static ssize_t ucma_bind_ip(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr);
+ mutex_unlock(&ctx->mutex);
+
ucma_put_ctx(ctx);
return ret;
}
@@ -681,7 +687,9 @@ static ssize_t ucma_bind(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -705,8 +713,10 @@ static ssize_t ucma_resolve_ip(struct ucma_file *file,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr,
(struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -731,8 +741,10 @@ static ssize_t ucma_resolve_addr(struct ucma_file *file,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr,
(struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -752,7 +764,9 @@ static ssize_t ucma_resolve_route(struct ucma_file *file,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_resolve_route(ctx->cm_id, cmd.timeout_ms);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -841,6 +855,7 @@ static ssize_t ucma_query_route(struct ucma_file *file,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
memset(&resp, 0, sizeof resp);
addr = (struct sockaddr *) &ctx->cm_id->route.addr.src_addr;
memcpy(&resp.src_addr, addr, addr->sa_family == AF_INET ?
@@ -864,6 +879,7 @@ static ssize_t ucma_query_route(struct ucma_file *file,
ucma_copy_iw_route(&resp, &ctx->cm_id->route);

out:
+ mutex_unlock(&ctx->mutex);
if (copy_to_user(u64_to_user_ptr(cmd.response),
&resp, sizeof(resp)))
ret = -EFAULT;
@@ -1014,6 +1030,7 @@ static ssize_t ucma_query(struct ucma_file *file,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
switch (cmd.option) {
case RDMA_USER_CM_QUERY_ADDR:
ret = ucma_query_addr(ctx, response, out_len);
@@ -1028,6 +1045,7 @@ static ssize_t ucma_query(struct ucma_file *file,
ret = -ENOSYS;
break;
}
+ mutex_unlock(&ctx->mutex);

ucma_put_ctx(ctx);
return ret;
@@ -1068,7 +1086,9 @@ static ssize_t ucma_connect(struct ucma_file *file, const char __user *inbuf,
return PTR_ERR(ctx);

ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param);
+ mutex_lock(&ctx->mutex);
ret = rdma_connect(ctx->cm_id, &conn_param);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -1089,7 +1109,9 @@ static ssize_t ucma_listen(struct ucma_file *file, const char __user *inbuf,

ctx->backlog = cmd.backlog > 0 && cmd.backlog < max_backlog ?
cmd.backlog : max_backlog;
+ mutex_lock(&ctx->mutex);
ret = rdma_listen(ctx->cm_id, ctx->backlog);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -1112,13 +1134,17 @@ static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
if (cmd.conn_param.valid) {
ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param);
mutex_lock(&file->mut);
+ mutex_lock(&ctx->mutex);
ret = __rdma_accept(ctx->cm_id, &conn_param, NULL);
+ mutex_unlock(&ctx->mutex);
if (!ret)
ctx->uid = cmd.uid;
mutex_unlock(&file->mut);
- } else
+ } else {
+ mutex_lock(&ctx->mutex);
ret = __rdma_accept(ctx->cm_id, NULL, NULL);
-
+ mutex_unlock(&ctx->mutex);
+ }
ucma_put_ctx(ctx);
return ret;
}
@@ -1137,7 +1163,9 @@ static ssize_t ucma_reject(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_reject(ctx->cm_id, cmd.private_data, cmd.private_data_len);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -1156,7 +1184,9 @@ static ssize_t ucma_disconnect(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
ret = rdma_disconnect(ctx->cm_id);
+ mutex_unlock(&ctx->mutex);
ucma_put_ctx(ctx);
return ret;
}
@@ -1187,7 +1217,9 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
resp.qp_attr_mask = 0;
memset(&qp_attr, 0, sizeof qp_attr);
qp_attr.qp_state = cmd.qp_state;
+ mutex_lock(&ctx->mutex);
ret = rdma_init_qp_attr(ctx->cm_id, &qp_attr, &resp.qp_attr_mask);
+ mutex_unlock(&ctx->mutex);
if (ret)
goto out;

@@ -1273,9 +1305,13 @@ static int ucma_set_ib_path(struct ucma_context *ctx,
struct sa_path_rec opa;

sa_convert_path_ib_to_opa(&opa, &sa_path);
+ mutex_lock(&ctx->mutex);
ret = rdma_set_ib_path(ctx->cm_id, &opa);
+ mutex_unlock(&ctx->mutex);
} else {
+ mutex_lock(&ctx->mutex);
ret = rdma_set_ib_path(ctx->cm_id, &sa_path);
+ mutex_unlock(&ctx->mutex);
}
if (ret)
return ret;
@@ -1308,7 +1344,9 @@ static int ucma_set_option_level(struct ucma_context *ctx, int level,

switch (level) {
case RDMA_OPTION_ID:
+ mutex_lock(&ctx->mutex);
ret = ucma_set_option_id(ctx, optname, optval, optlen);
+ mutex_unlock(&ctx->mutex);
break;
case RDMA_OPTION_IB:
ret = ucma_set_option_ib(ctx, optname, optval, optlen);
@@ -1368,8 +1406,10 @@ static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);

+ mutex_lock(&ctx->mutex);
if (ctx->cm_id->device)
ret = rdma_notify(ctx->cm_id, (enum ib_event_type)cmd.event);
+ mutex_unlock(&ctx->mutex);

ucma_put_ctx(ctx);
return ret;
@@ -1412,8 +1452,10 @@ static ssize_t ucma_process_join(struct ucma_file *file,
mc->join_state = join_state;
mc->uid = cmd->uid;
memcpy(&mc->addr, addr, cmd->addr_size);
+ mutex_lock(&ctx->mutex);
ret = rdma_join_multicast(ctx->cm_id, (struct sockaddr *)&mc->addr,
join_state, mc);
+ mutex_unlock(&ctx->mutex);
if (ret)
goto err2;

@@ -1513,7 +1555,10 @@ static ssize_t ucma_leave_multicast(struct ucma_file *file,
goto out;
}

+ mutex_lock(&mc->ctx->mutex);
rdma_leave_multicast(mc->ctx->cm_id, (struct sockaddr *) &mc->addr);
+ mutex_unlock(&mc->ctx->mutex);
+
mutex_lock(&mc->ctx->file->mut);
ucma_cleanup_mc_events(mc);
list_del(&mc->list);
--
2.20.1

Next message: Sasha Levin: "[PATCH AUTOSEL 5.5 071/121] drm/omap: dss: Cleanup DSS ports on initialisation failure"
Previous message: Sasha Levin: "[PATCH AUTOSEL 5.5 062/121] drm/amdgpu: fix parentheses in amdgpu_vm_update_ptes"
In reply to: Sasha Levin: "[PATCH AUTOSEL 5.5 062/121] drm/amdgpu: fix parentheses in amdgpu_vm_update_ptes"
Next in thread: Sasha Levin: "[PATCH AUTOSEL 5.5 071/121] drm/omap: dss: Cleanup DSS ports on initialisation failure"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]