Skip to contentSkip to conversation

RDMA/ucma: fix a kernel-infoleak in ucma_init_qp_attr()

PATCH Superseded Dan Carpenter 4 participants 4 messages Feb 3, 2022 InfiniBand

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 RDMA/ucma: fix a kernel-infoleak in ucma_init_qp_attr()

From: Haimin Zhang <[email protected]>

The ib_copy_ah_attr_to_user() function only initializes "resp.grh" if the "resp.is_global" flag is set. Unfortunately, this data is copied to the user and copying uninitialized stack data to the user is an information leak. Zero out the whole struct to be safe.

Fixes 4ba66093bdc6 ("IB/core: Check for global flag when using ah_attr")

---
Resending through the regular lists.

I added parentheses around the sizeof to make checkpatch happy. s/sizeof resp/sizeof(resp)/.

 drivers/infiniband/core/ucma.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/infiniband/core/ucma.c +1 -1
index 9d6ac9dff39a..91485f13d842 100644
@@ -1232,7 +1232,7 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
-	resp.qp_attr_mask = 0;
+	memset(&resp, 0, sizeof(resp));
 	memset(&qp_attr, 0, sizeof qp_attr);
 	qp_attr.qp_state = cmd.qp_state;
 	mutex_lock(&ctx->mutex);
Signature · 1 line
2.20.1
#1 lore raw
Quoted 11 lines · Dan Carpenter

From: Haimin Zhang <[email protected]>

The ib_copy_ah_attr_to_user() function only initializes "resp.grh" if the "resp.is_global" flag is set. Unfortunately, this data is copied to the user and copying uninitialized stack data to the user is an information leak. Zero out the whole struct to be safe.

Fixes 4ba66093bdc6 ("IB/core: Check for global flag when using ah_attr")
Quoted 8 lines

---
Resending through the regular lists.

I added parentheses around the sizeof to make checkpatch happy. s/sizeof resp/sizeof(resp)/.

 drivers/infiniband/core/ucma.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

The change is ok, but I prefer to initialize to zero as early as possible.

drivers/infiniband/core/ucma.c +2 -4
index 2b72c4fa9550..6d801ed2e46b 100644
@@ -1211,9 +1211,9 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
                                 int in_len, int out_len)
 {
        struct rdma_ucm_init_qp_attr cmd;
-       struct ib_uverbs_qp_attr resp;
+       struct ib_uverbs_qp_attr resp = {};
        struct ucma_context *ctx;
-       struct ib_qp_attr qp_attr;
+       struct ib_qp_attr qp_attr = {};
        int ret;
 
        if (out_len < sizeof(resp))
@@ -1229,8 +1229,6 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
        if (IS_ERR(ctx))
                return PTR_ERR(ctx);
 
-       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);
Quoted 17 lines
drivers/infiniband/core/ucma.c +1 -1
index 9d6ac9dff39a..91485f13d842 100644
@@ -1232,7 +1232,7 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
-	resp.qp_attr_mask = 0;
+	memset(&resp, 0, sizeof(resp));
 	memset(&qp_attr, 0, sizeof qp_attr);
 	qp_attr.qp_state = cmd.qp_state;
 	mutex_lock(&ctx->mutex);
Signature · 1 line
2.20.1
#2 lore raw
Quoted 28 lines · Leon Romanovsky
Quoted 11 lines · Dan Carpenter

From: Haimin Zhang <[email protected]>

The ib_copy_ah_attr_to_user() function only initializes "resp.grh" if the "resp.is_global" flag is set. Unfortunately, this data is copied to the user and copying uninitialized stack data to the user is an information leak. Zero out the whole struct to be safe.

Fixes 4ba66093bdc6 ("IB/core: Check for global flag when using ah_attr")
Quoted 8 lines

---
Resending through the regular lists.

I added parentheses around the sizeof to make checkpatch happy. s/sizeof resp/sizeof(resp)/.

 drivers/infiniband/core/ucma.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

The change is ok, but I prefer to initialize to zero as early as possible.

drivers/infiniband/core/ucma.c +0 -0
index 2b72c4fa9550..6d801ed2e46b 100644

drivers/infiniband/core/ucma.c static ssize_t ucma_init_qp_attr(struct ucma_file *file, · line 1216

Quoted 8 lines
drivers/infiniband/core/ucma.c +2 -2
@@ -1211,9 +1211,9 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
                                 int in_len, int out_len)
 {
        struct rdma_ucm_init_qp_attr cmd;
-       struct ib_uverbs_qp_attr resp;
+       struct ib_uverbs_qp_attr resp = {};
        struct ucma_context *ctx;
-       struct ib_qp_attr qp_attr;
+       struct ib_qp_attr qp_attr = {};

Will that catch all of the padding in the structure? This seems to come up a lot and I never remember...

thanks,

greg k-h

#3 lore raw
Quoted 23 lines · Greg KH
Quoted 22 lines · Leon Romanovsky
Quoted 11 lines · Dan Carpenter

From: Haimin Zhang <[email protected]>

The ib_copy_ah_attr_to_user() function only initializes "resp.grh" if the "resp.is_global" flag is set. Unfortunately, this data is copied to the user and copying uninitialized stack data to the user is an information leak. Zero out the whole struct to be safe.

Fixes 4ba66093bdc6 ("IB/core: Check for global flag when using ah_attr")
Quoted 7 lines

Resending through the regular lists.

I added parentheses around the sizeof to make checkpatch happy. s/sizeof resp/sizeof(resp)/.

 drivers/infiniband/core/ucma.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

The change is ok, but I prefer to initialize to zero as early as possible.

Quoted 15 lines

drivers/infiniband/core/ucma.c static ssize_t ucma_init_qp_attr(struct ucma_file *file, · line 1216

Quoted 12 lines
drivers/infiniband/core/ucma.c +3 -2
index 2b72c4fa9550..6d801ed2e46b 100644
+++ b/drivers/infiniband/core/ucma.c
@@ -1211,9 +1211,9 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
                                 int in_len, int out_len)
 {
        struct rdma_ucm_init_qp_attr cmd;
-       struct ib_uverbs_qp_attr resp;
+       struct ib_uverbs_qp_attr resp = {};
        struct ucma_context *ctx;
-       struct ib_qp_attr qp_attr;
+       struct ib_qp_attr qp_attr = {};

Will that catch all of the padding in the structure? This seems to come up a lot and I never remember...

Yes, last time you asked we went over it.

Jason