create_flow/destroy_flow fixes for v3.13

PATCH In review Yann Droneaud 4 participants 17 messages Dec 17, 2013 – Dec 19, 2013 5 messages not in the archive InfiniBand

lore

Share & tools

lore

Yann Droneaud

author

2013-12-17 09:58 UTC #0 lore raw

v4 cover create_flow/destroy_flow fixes for v3.13

Hi Roland,

Please find the fourth revision of a patchset against create_flow/destroy_flow and associated extended command scheme.

These are fixes that *must* be applied before making the new uverbs widely available in v3.13:

- patches 1, 2 and 3 address a warning reported by sparse.
- patch 4 handle an uncommon type of extended command.
- patch 5, 6 and 7 ensure that commands will be extensible:
- one patch add a missing check of comp_mask;
- the other patches add checks on reserved fields.
- patch 8 fix an error path which was not setting the error code.
- patch 9 ensure the response buffer of an extended command is a valid memory region for userspace. Doing the check earlier, on the *full* response buffer, might be safer.
- patch 10 prevent out of bound access and underflow.

Please review, test and apply for v3.13.

Changes from patchset v3 [1]:

- add a patch to hold 'response' pointer to remove the redundant cast introduced by patch to remove implicit cast from INIT_UDATA()
- rework some commit messages to address review made by Roland.

Changes from patchset v2 [2]:

- add a patch to check the input length in extended uverbs, create_flow and destroy_flow.

Changes from patchset v1 [3]:

- add patch to verify early response buffer using access_ok(), just like vfs_read().

Regards.

[1] [PATCHv3 for-3.13 0/9] create_flow/destroy_flow fixes for v3.13

    http://marc.info/?i=cover.1386798254.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

[2] [PATCH v2 for v3.13 0/8] create_flow/destroy_flow fixes for v3.13

    http://marc.info/?i=cover.1385981934.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

[3] [PATCH for v3.13 0/7] create_flow/destroy_flow fixes for v3.13

    http://marc.info/?i=cover.1385501822.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Yann Droneaud (10):
IB/uverbs: move cast from u64 to void __user pointer to it's own
variable
IB/core: const'ify inbuf in struct ib_udata
IB/uverbs: remove implicit cast in INIT_UDATA()
IB/uverbs: set outbuf to NULL when no core response space is provided IB/uverbs: check reserved field in extended command header IB/uverbs: check comp_mask in destroy_flow
IB/uverbs: check reserved fields in create_flow
IB/uverbs: set error code when fail to consume all flow_spec items IB/uverbs: check access to userspace response buffer in extended
command
IB/uverbs: check input length in flow steering uverbs

 drivers/infiniband/core/uverbs.h      |  12 +--
 drivers/infiniband/core/uverbs_cmd.c  | 175 +++++++++++++++++++++++-----------
 drivers/infiniband/core/uverbs_main.c |  21 +++-
 include/rdma/ib_verbs.h               |   2 +-
 4 files changed, 144 insertions(+), 66 deletions(-)

Signature · 6 lines

1.8.4.2

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

A message that is not in the archive (sent off-list or not yet received): cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Yann Droneaud

author

2013-12-17 09:58 UTC #2 lore raw

PATCH v4 1/10 IB/uverbs: move cast from u64 to void __user pointer to it's own variable

Use a dedicated variable to hold address of the response buffer after 'conversion' from u64 to void __user *, so that this value could be used for INIT_UDATA() and copy_to_user(), reducing the visual clutter introduced by the cast.

This variable will be used when implicit cast will be removed from INIT_UDATA() macro, which is required in order to remove a sparse warning.

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs_cmd.c | 158 +++++++++++++++++++++++------------
 1 file changed, 104 insertions(+), 54 deletions(-)

drivers/infiniband/core/uverbs_cmd.c +104 ~~-54~~

index 65f6e7dc380c..971e16c970b9 100644
@@ -285,6 +285,7 @@ ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_get_context      cmd;
 	struct ib_uverbs_get_context_resp resp;
+	char __user			 *response;
 	struct ib_udata                   udata;
 	struct ib_device                 *ibdev = file->device->ib_dev;
 	struct ib_ucontext		 *ucontext;

@@ -297,6 +298,8 @@ ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	mutex_lock(&file->mutex);
 
 	if (file->ucontext) {

@@ -305,7 +308,7 @@ ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file,
 	}
 
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	ucontext = ibdev->alloc_ucontext(ibdev, &udata);

@@ -339,8 +342,7 @@ ssize_t ib_uverbs_get_context(struct ib_uverbs_file *file,
 		goto err_fd;
 	}
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_file;
 	}

@@ -383,6 +385,7 @@ ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_query_device      cmd;
 	struct ib_uverbs_query_device_resp resp;
+	char __user			  *response;
 	struct ib_device_attr              attr;
 	int                                ret;

@@ -392,6 +395,8 @@ ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	ret = ib_query_device(file->device->ib_dev, &attr);
 	if (ret)
 		return ret;

@@ -439,8 +444,7 @@ ssize_t ib_uverbs_query_device(struct ib_uverbs_file *file,
 	resp.local_ca_ack_delay        = attr.local_ca_ack_delay;
 	resp.phys_port_cnt	       = file->device->ib_dev->phys_port_cnt;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		return -EFAULT;
 
 	return in_len;

@@ -452,6 +456,7 @@ ssize_t ib_uverbs_query_port(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_query_port      cmd;
 	struct ib_uverbs_query_port_resp resp;
+	char __user			*response;
 	struct ib_port_attr              attr;
 	int                              ret;

@@ -461,6 +466,8 @@ ssize_t ib_uverbs_query_port(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	ret = ib_query_port(file->device->ib_dev, cmd.port_num, &attr);
 	if (ret)
 		return ret;

@@ -489,8 +496,7 @@ ssize_t ib_uverbs_query_port(struct ib_uverbs_file *file,
 	resp.link_layer      = rdma_port_get_link_layer(file->device->ib_dev,
 							cmd.port_num);
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		return -EFAULT;
 
 	return in_len;

@@ -502,6 +508,7 @@ ssize_t ib_uverbs_alloc_pd(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_alloc_pd      cmd;
 	struct ib_uverbs_alloc_pd_resp resp;
+	char __user		      *response;
 	struct ib_udata                udata;
 	struct ib_uobject             *uobj;
 	struct ib_pd                  *pd;

@@ -513,8 +520,10 @@ ssize_t ib_uverbs_alloc_pd(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	uobj = kmalloc(sizeof *uobj, GFP_KERNEL);

@@ -543,8 +552,7 @@ ssize_t ib_uverbs_alloc_pd(struct ib_uverbs_file *file,
 	memset(&resp, 0, sizeof resp);
 	resp.pd_handle = uobj->id;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -696,6 +704,7 @@ ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_open_xrcd	cmd;
 	struct ib_uverbs_open_xrcd_resp	resp;
+	char __user		       *response;
 	struct ib_udata			udata;
 	struct ib_uxrcd_object         *obj;
 	struct ib_xrcd                 *xrcd = NULL;

@@ -710,8 +719,10 @@ ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof  resp);
 
 	mutex_lock(&file->device->xrcd_tree_mutex);

@@ -783,8 +794,7 @@ ssize_t ib_uverbs_open_xrcd(struct ib_uverbs_file *file,
 		atomic_inc(&xrcd->usecnt);
 	}
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -910,6 +920,7 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_reg_mr      cmd;
 	struct ib_uverbs_reg_mr_resp resp;
+	char __user		    *response;
 	struct ib_udata              udata;
 	struct ib_uobject           *uobj;
 	struct ib_pd                *pd;

@@ -922,8 +933,10 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	if ((cmd.start & ~PAGE_MASK) != (cmd.hca_va & ~PAGE_MASK))

@@ -969,8 +982,7 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 	resp.rkey      = mr->rkey;
 	resp.mr_handle = uobj->id;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -1045,6 +1057,7 @@ ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_alloc_mw      cmd;
 	struct ib_uverbs_alloc_mw_resp resp;
+	char __user		      *response;
 	struct ib_uobject             *uobj;
 	struct ib_pd                  *pd;
 	struct ib_mw                  *mw;

@@ -1056,6 +1069,8 @@ ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof(cmd)))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	uobj = kmalloc(sizeof(*uobj), GFP_KERNEL);
 	if (!uobj)
 		return -ENOMEM;

@@ -1089,8 +1104,7 @@ ssize_t ib_uverbs_alloc_mw(struct ib_uverbs_file *file,
 	resp.rkey      = mw->rkey;
 	resp.mw_handle = uobj->id;
 
-	if (copy_to_user((void __user *)(unsigned long)cmd.response,
-			 &resp, sizeof(resp))) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -1165,6 +1179,7 @@ ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_create_comp_channel	   cmd;
 	struct ib_uverbs_create_comp_channel_resp  resp;
+	char __user				  *response;
 	struct file				  *filp;
 	int ret;

@@ -1174,6 +1189,8 @@ ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	ret = get_unused_fd_flags(O_CLOEXEC);
 	if (ret < 0)
 		return ret;

@@ -1185,8 +1202,7 @@ ssize_t ib_uverbs_create_comp_channel(struct ib_uverbs_file *file,
 		return PTR_ERR(filp);
 	}
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		put_unused_fd(resp.fd);
 		fput(filp);
 		return -EFAULT;

@@ -1202,6 +1218,7 @@ ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_create_cq      cmd;
 	struct ib_uverbs_create_cq_resp resp;
+	char __user		       *response;
 	struct ib_udata                 udata;
 	struct ib_ucq_object           *obj;
 	struct ib_uverbs_event_file    *ev_file = NULL;

@@ -1214,8 +1231,10 @@ ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	if (cmd.comp_vector >= file->device->num_comp_vectors)

@@ -1266,8 +1285,7 @@ ssize_t ib_uverbs_create_cq(struct ib_uverbs_file *file,
 	resp.cq_handle = obj->uobject.id;
 	resp.cqe       = cq->cqe;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -1303,6 +1321,7 @@ ssize_t ib_uverbs_resize_cq(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_resize_cq	cmd;
 	struct ib_uverbs_resize_cq_resp	resp;
+	char __user		       *response;
 	struct ib_udata                 udata;
 	struct ib_cq			*cq;
 	int				ret = -EINVAL;

@@ -1310,8 +1329,10 @@ ssize_t ib_uverbs_resize_cq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	cq = idr_read_cq(cmd.cq_handle, file->ucontext, 0);

@@ -1324,8 +1345,7 @@ ssize_t ib_uverbs_resize_cq(struct ib_uverbs_file *file,
 
 	resp.cqe = cq->cqe;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp.cqe))
+	if (copy_to_user(response, &resp, sizeof(resp.cqe)))
 		ret = -EFAULT;
 
 out:

@@ -1439,6 +1459,7 @@ ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_destroy_cq      cmd;
 	struct ib_uverbs_destroy_cq_resp resp;
+	char __user			*response;
 	struct ib_uobject		*uobj;
 	struct ib_cq               	*cq;
 	struct ib_ucq_object        	*obj;

@@ -1448,6 +1469,8 @@ ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	uobj = idr_write_uobj(&ib_uverbs_cq_idr, cmd.cq_handle, file->ucontext);
 	if (!uobj)
 		return -EINVAL;

@@ -1478,8 +1501,7 @@ ssize_t ib_uverbs_destroy_cq(struct ib_uverbs_file *file,
 
 	put_uobj(uobj);
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		return -EFAULT;
 
 	return in_len;

@@ -1491,6 +1513,7 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_create_qp      cmd;
 	struct ib_uverbs_create_qp_resp resp;
+	char __user		       *response;
 	struct ib_udata                 udata;
 	struct ib_uqp_object           *obj;
 	struct ib_device	       *device;

@@ -1512,8 +1535,10 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 	if (cmd.qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW))
 		return -EPERM;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	obj = kzalloc(sizeof *obj, GFP_KERNEL);

@@ -1626,8 +1651,7 @@ ssize_t ib_uverbs_create_qp(struct ib_uverbs_file *file,
 	resp.max_send_wr     = attr.cap.max_send_wr;
 	resp.max_inline_data = attr.cap.max_inline_data;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -1685,6 +1709,7 @@ ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_open_qp        cmd;
 	struct ib_uverbs_create_qp_resp resp;
+	char __user		       *response;
 	struct ib_udata                 udata;
 	struct ib_uqp_object           *obj;
 	struct ib_xrcd		       *xrcd;

@@ -1699,8 +1724,10 @@ ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	obj = kmalloc(sizeof *obj, GFP_KERNEL);

@@ -1742,8 +1769,7 @@ ssize_t ib_uverbs_open_qp(struct ib_uverbs_file *file,
 	resp.qpn       = qp->qp_num;
 	resp.qp_handle = obj->uevent.uobject.id;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_remove;
 	}

@@ -1780,6 +1806,7 @@ ssize_t ib_uverbs_query_qp(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_query_qp      cmd;
 	struct ib_uverbs_query_qp_resp resp;
+	char __user		       *response;
 	struct ib_qp                   *qp;
 	struct ib_qp_attr              *attr;
 	struct ib_qp_init_attr         *init_attr;

@@ -1788,6 +1815,8 @@ ssize_t ib_uverbs_query_qp(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	attr      = kmalloc(sizeof *attr, GFP_KERNEL);
 	init_attr = kmalloc(sizeof *init_attr, GFP_KERNEL);
 	if (!attr || !init_attr) {

@@ -1863,8 +1892,7 @@ ssize_t ib_uverbs_query_qp(struct ib_uverbs_file *file,
 	resp.max_inline_data        = init_attr->cap.max_inline_data;
 	resp.sq_sig_all             = init_attr->sq_sig_type == IB_SIGNAL_ALL_WR;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		ret = -EFAULT;
 
 out:

@@ -1986,6 +2014,7 @@ ssize_t ib_uverbs_destroy_qp(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_destroy_qp      cmd;
 	struct ib_uverbs_destroy_qp_resp resp;
+	char __user			*response;
 	struct ib_uobject		*uobj;
 	struct ib_qp               	*qp;
 	struct ib_uqp_object        	*obj;

@@ -1994,6 +2023,8 @@ ssize_t ib_uverbs_destroy_qp(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	memset(&resp, 0, sizeof resp);
 
 	uobj = idr_write_uobj(&ib_uverbs_qp_idr, cmd.qp_handle, file->ucontext);

@@ -2031,8 +2062,7 @@ ssize_t ib_uverbs_destroy_qp(struct ib_uverbs_file *file,
 
 	put_uobj(uobj);
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		return -EFAULT;
 
 	return in_len;

@@ -2044,6 +2074,7 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_post_send      cmd;
 	struct ib_uverbs_post_send_resp resp;
+	char __user		       *response;
 	struct ib_uverbs_send_wr       *user_wr;
 	struct ib_send_wr              *wr = NULL, *last, *next, *bad_wr;
 	struct ib_qp                   *qp;

@@ -2061,6 +2092,8 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 	if (cmd.wqe_size < sizeof (struct ib_uverbs_send_wr))
 		return -EINVAL;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	user_wr = kmalloc(cmd.wqe_size, GFP_KERNEL);
 	if (!user_wr)
 		return -ENOMEM;

@@ -2176,8 +2209,7 @@ ssize_t ib_uverbs_post_send(struct ib_uverbs_file *file,
 				break;
 		}
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		ret = -EFAULT;
 
 out_put:

@@ -2288,6 +2320,7 @@ ssize_t ib_uverbs_post_recv(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_post_recv      cmd;
 	struct ib_uverbs_post_recv_resp resp;
+	char __user		       *response;
 	struct ib_recv_wr              *wr, *next, *bad_wr;
 	struct ib_qp                   *qp;
 	ssize_t                         ret = -EINVAL;

@@ -2295,6 +2328,8 @@ ssize_t ib_uverbs_post_recv(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	wr = ib_uverbs_unmarshall_recv(buf + sizeof cmd,
 				       in_len - sizeof cmd, cmd.wr_count,
 				       cmd.sge_count, cmd.wqe_size);

@@ -2317,8 +2352,7 @@ ssize_t ib_uverbs_post_recv(struct ib_uverbs_file *file,
 				break;
 		}
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		ret = -EFAULT;
 
 out:

@@ -2337,6 +2371,7 @@ ssize_t ib_uverbs_post_srq_recv(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_post_srq_recv      cmd;
 	struct ib_uverbs_post_srq_recv_resp resp;
+	char __user			   *response;
 	struct ib_recv_wr                  *wr, *next, *bad_wr;
 	struct ib_srq                      *srq;
 	ssize_t                             ret = -EINVAL;

@@ -2344,6 +2379,8 @@ ssize_t ib_uverbs_post_srq_recv(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	wr = ib_uverbs_unmarshall_recv(buf + sizeof cmd,
 				       in_len - sizeof cmd, cmd.wr_count,
 				       cmd.sge_count, cmd.wqe_size);

@@ -2366,8 +2403,7 @@ ssize_t ib_uverbs_post_srq_recv(struct ib_uverbs_file *file,
 				break;
 		}
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		ret = -EFAULT;
 
 out:

@@ -2386,6 +2422,7 @@ ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_create_ah	 cmd;
 	struct ib_uverbs_create_ah_resp	 resp;
+	char __user			*response;
 	struct ib_uobject		*uobj;
 	struct ib_pd			*pd;
 	struct ib_ah			*ah;

@@ -2398,6 +2435,8 @@ ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	uobj = kmalloc(sizeof *uobj, GFP_KERNEL);
 	if (!uobj)
 		return -ENOMEM;

@@ -2438,8 +2477,7 @@ ssize_t ib_uverbs_create_ah(struct ib_uverbs_file *file,
 
 	resp.ah_handle = uobj->id;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -2823,6 +2861,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 				struct ib_udata *udata)
 {
 	struct ib_uverbs_create_srq_resp resp;
+	char __user			*response;
 	struct ib_usrq_object           *obj;
 	struct ib_pd                    *pd;
 	struct ib_srq                   *srq;

@@ -2830,6 +2869,8 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 	struct ib_srq_init_attr          attr;
 	int ret;
 
+	response = (void __user *)(unsigned long)cmd->response;
+
 	obj = kmalloc(sizeof *obj, GFP_KERNEL);
 	if (!obj)
 		return -ENOMEM;

@@ -2905,8 +2946,7 @@ static int __uverbs_create_xsrq(struct ib_uverbs_file *file,
 	if (cmd->srq_type == IB_SRQT_XRC)
 		resp.srqn = srq->ext.xrc.srq_num;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd->response,
-			 &resp, sizeof resp)) {
+	if (copy_to_user(response, &resp, sizeof(resp))) {
 		ret = -EFAULT;
 		goto err_copy;
 	}

@@ -2958,6 +2998,7 @@ ssize_t ib_uverbs_create_srq(struct ib_uverbs_file *file,
 	struct ib_uverbs_create_srq      cmd;
 	struct ib_uverbs_create_xsrq     xcmd;
 	struct ib_uverbs_create_srq_resp resp;
+	char __user			*response;
 	struct ib_udata                  udata;
 	int ret;

@@ -2967,6 +3008,8 @@ ssize_t ib_uverbs_create_srq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	xcmd.response	 = cmd.response;
 	xcmd.user_handle = cmd.user_handle;
 	xcmd.srq_type	 = IB_SRQT_BASIC;

@@ -2976,7 +3019,7 @@ ssize_t ib_uverbs_create_srq(struct ib_uverbs_file *file,
 	xcmd.srq_limit	 = cmd.srq_limit;
 
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	ret = __uverbs_create_xsrq(file, &xcmd, &udata);

@@ -2991,6 +3034,7 @@ ssize_t ib_uverbs_create_xsrq(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_create_xsrq     cmd;
 	struct ib_uverbs_create_srq_resp resp;
+	char __user			*response;
 	struct ib_udata                  udata;
 	int ret;

@@ -3000,8 +3044,10 @@ ssize_t ib_uverbs_create_xsrq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	INIT_UDATA(&udata, buf + sizeof cmd,
-		   (unsigned long) cmd.response + sizeof resp,
+		   response + sizeof resp,
 		   in_len - sizeof cmd, out_len - sizeof resp);
 
 	ret = __uverbs_create_xsrq(file, &cmd, &udata);

@@ -3047,6 +3093,7 @@ ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_query_srq      cmd;
 	struct ib_uverbs_query_srq_resp resp;
+	char __user		       *response;
 	struct ib_srq_attr              attr;
 	struct ib_srq                   *srq;
 	int                             ret;

@@ -3057,6 +3104,8 @@ ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	srq = idr_read_srq(cmd.srq_handle, file->ucontext);
 	if (!srq)
 		return -EINVAL;

@@ -3074,8 +3123,7 @@ ssize_t ib_uverbs_query_srq(struct ib_uverbs_file *file,
 	resp.max_sge   = attr.max_sge;
 	resp.srq_limit = attr.srq_limit;
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		return -EFAULT;
 
 	return in_len;

@@ -3087,6 +3135,7 @@ ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file,
 {
 	struct ib_uverbs_destroy_srq      cmd;
 	struct ib_uverbs_destroy_srq_resp resp;
+	char __user			 *response;
 	struct ib_uobject		 *uobj;
 	struct ib_srq               	 *srq;
 	struct ib_uevent_object        	 *obj;

@@ -3097,6 +3146,8 @@ ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file,
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;
 
+	response = (void __user *)(unsigned long)cmd.response;
+
 	uobj = idr_write_uobj(&ib_uverbs_srq_idr, cmd.srq_handle, file->ucontext);
 	if (!uobj)
 		return -EINVAL;

@@ -3131,8 +3182,7 @@ ssize_t ib_uverbs_destroy_srq(struct ib_uverbs_file *file,
 
 	put_uobj(uobj);
 
-	if (copy_to_user((void __user *) (unsigned long) cmd.response,
-			 &resp, sizeof resp))
+	if (copy_to_user(response, &resp, sizeof(resp)))
 		ret = -EFAULT;
 
 	return ret ? ret : in_len;

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #3 lore raw

PATCH v4 2/10 IB/core: const'ify inbuf in struct ib_udata

In reply to a message not in the archive

Userspace input buffer is not modified by kernel, let it be 'const'.

It's a prerequisite to remove the implicit cast from INIT_UDATA(), which is needed to remove a sparse warning.

Without this change on struct ib_udata, compiler will complain when input buffer (const) will be stored in inbuf (non-const).

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs.h | 2 +-
 include/rdma/ib_verbs.h          | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

drivers/infiniband/core/uverbs.h +1 -1

index bdc842e9faef..9879568aed8c 100644
@@ -49,7 +49,7 @@
 
 #define INIT_UDATA(udata, ibuf, obuf, ilen, olen)			\
 	do {								\
-		(udata)->inbuf  = (void __user *) (ibuf);		\
+		(udata)->inbuf  = (const void __user *) (ibuf);		\
 		(udata)->outbuf = (void __user *) (obuf);		\
 		(udata)->inlen  = (ilen);				\
 		(udata)->outlen = (olen);				\

include/rdma/ib_verbs.h +1 -1

index 979874c627ee..61e1935c91b1 100644
@@ -978,7 +978,7 @@ struct ib_uobject {
 };
 
 struct ib_udata {
-	void __user *inbuf;
+	const void __user *inbuf;
 	void __user *outbuf;
 	size_t       inlen;
 	size_t       outlen;

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #4 lore raw

PATCH v4 3/10 IB/uverbs: remove implicit cast in INIT_UDATA()

In reply to a message not in the archive

Currently, INIT_UDATA() does an implicit cast to a pointer, so that 'response' address, eg. output buffer, can be used as is to initialize a struct ib_udata:

        do {                                                    \
                (udata)->inbuf  = (void __user *) (ibuf);       \
                (udata)->outbuf = (void __user *) (obuf);       \
                (udata)->inlen  = (ilen);                       \
                (udata)->outlen = (olen);                       \
        } while (0)

...

INIT_UDATA(&udata, buf + sizeof cmd,
(unsigned long) cmd.response + sizeof resp, in_len - sizeof cmd, out_len - sizeof resp);

...

Unfortunately, sparse reports an error if literal 0 is used to initialize inbuf or outbuf, for example in:

        INIT_UDATA(&ucore,
                   (hdr.in_words) ? buf : 0,
                   (unsigned long)ex_hdr.response,
                   hdr.in_words * 8,
                   hdr.out_words * 8);

It was reported by kbuild test robot in message[1]:

From: kbuild test robot <[email protected]>
Subject: "drivers/infiniband/core/uverbs_main.c:683:17:
sparse: Using plain integer as NULL pointer",
Message-Id: <528b3984.SVGs20ZWpcuR/Jls%[email protected]>

Replacing 0 by NULL in ib_uverbs_write() might turn off sparse warning, but it trigger a GCC warning:

drivers/infiniband/core/uverbs_main.c: In function 'ib_uverbs_write':
drivers/infiniband/core/uverbs_main.c:685:200: warning: pointer/integer type mismatch in conditional expression [enabled by default]

So fixing the sparse warning require broad changes in uverbs.

By removing the implicit cast INIT_UDATA(), this patch fixes the warnings reported by sparse and allows the compiler to report a warning in case a plain integer get used to initialize an udata pointer.

This patch requires
- calls to INIT_UDATA() to be modified to pass a pointer instead of an unsigned long,
- struct ib_udata to be modified to have a const void __user *inbuf field[2], otherwise compiler will report warnings regarding const to non const conversion:

drivers/infiniband/core/uverbs_main.c: In function ‘ib_uverbs_write’:
drivers/infiniband/core/uverbs_main.c:682:24: attention : assignment discards ‘const’ qualifier from pointer target type [enabled by default] drivers/infiniband/core/uverbs_main.c:688:22: attention : assignment discards ‘const’ qualifier from pointer target type [enabled by default] drivers/infiniband/core/uverbs_cmd.c: In function ‘ib_uverbs_get_context’:
drivers/infiniband/core/uverbs_cmd.c:307:23: attention : assignment discards ‘const’ qualifier from pointer target type [enabled by default] drivers/infiniband/core/uverbs_cmd.c: In function ‘ib_uverbs_alloc_pd’:
drivers/infiniband/core/uverbs_cmd.c:516:23: attention : assignment discards ‘const’ qualifier from pointer target type [enabled by default]
...

[1] https://lists.01.org/pipermail/kbuild-all/2013-November/002120.html

[2] https://patchwork.kernel.org/patch/2846202/

    http://marc.info/?i=3050a98379b4342ea59d59aeaf1ce162171df928.1376847403.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs.h      | 12 ++++++------
 drivers/infiniband/core/uverbs_main.c | 13 ++++++++-----
 2 files changed, 14 insertions(+), 11 deletions(-)

drivers/infiniband/core/uverbs.h +6 -6

index 9879568aed8c..0dca1975d59d 100644
@@ -47,12 +47,12 @@
 #include <rdma/ib_umem.h>
 #include <rdma/ib_user_verbs.h>
 
-#define INIT_UDATA(udata, ibuf, obuf, ilen, olen)			\
-	do {								\
-		(udata)->inbuf  = (const void __user *) (ibuf);		\
-		(udata)->outbuf = (void __user *) (obuf);		\
-		(udata)->inlen  = (ilen);				\
-		(udata)->outlen = (olen);				\
+#define INIT_UDATA(udata, ibuf, obuf, ilen, olen)	\
+	do {						\
+		(udata)->inbuf  = (ibuf);		\
+		(udata)->outbuf = (obuf);		\
+		(udata)->inlen  = (ilen);		\
+		(udata)->outlen = (olen);		\
 	} while (0)
 
 /*

drivers/infiniband/core/uverbs_main.c +8 -5

index 34386943ebcf..14d864371050 100644
@@ -635,6 +635,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		__u32 command;
 
 		struct ib_uverbs_ex_cmd_hdr ex_hdr;
+		char __user *response;
 		struct ib_udata ucore;
 		struct ib_udata uhw;
 		int err;

@@ -668,7 +669,9 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count)
 			return -EINVAL;
 
-		if (ex_hdr.response) {
+		response = (char __user *)(unsigned long)ex_hdr.response;
+
+		if (response) {
 			if (!hdr.out_words && !ex_hdr.provider_out_words)
 				return -EINVAL;
 		} else {

@@ -677,14 +680,14 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		}
 
 		INIT_UDATA(&ucore,
-			   (hdr.in_words) ? buf : 0,
-			   (unsigned long)ex_hdr.response,
+			   (hdr.in_words) ? buf : NULL,
+			   response,
 			   hdr.in_words * 8,
 			   hdr.out_words * 8);
 
 		INIT_UDATA(&uhw,
-			   (ex_hdr.provider_in_words) ? buf + ucore.inlen : 0,
-			   (ex_hdr.provider_out_words) ? (unsigned long)ex_hdr.response + ucore.outlen : 0,
+			   (ex_hdr.provider_in_words) ? buf + ucore.inlen : NULL,
+			   (ex_hdr.provider_out_words) ? response + ucore.outlen : NULL,
 			   ex_hdr.provider_in_words * 8,
 			   ex_hdr.provider_out_words * 8);

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #5 lore raw

PATCH v4 4/10 IB/uverbs: set outbuf to NULL when no core response space is provided

In reply to a message not in the archive

In the currently uncommon case of core (eg. uverbs) response space being omitted, but hw (eg. provider) response space being available, outbuf get defined to "response" while it must be NULL.

This patch takes care of setting ucore->outbuf to NULL if hdr.out_words is equal to 0.

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

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

drivers/infiniband/core/uverbs_main.c +1 -1

index 14d864371050..6c4fc6338b26 100644
@@ -681,7 +681,7 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 
 		INIT_UDATA(&ucore,
 			   (hdr.in_words) ? buf : NULL,
-			   response,
+			   (hdr.out_words) ? response : NULL,
 			   hdr.in_words * 8,
 			   hdr.out_words * 8);

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #6 lore raw

PATCH v4 5/10 IB/uverbs: check reserved field in extended command header

In reply to a message not in the archive

As noted by Daniel Vetter in its article "Botching up ioctls"[1]

"Check *all* unused fields and flags and all the padding for
whether it's 0, and reject the ioctl if that's not the case. Otherwise your nice plan for future extensions is going right down the gutters since someone *will* submit an ioctl struct with random stack garbage in the yet unused parts. Which then bakes in the ABI that those fields can never be used for anything else but garbage."

It's important to ensure that reserved fields are set to known value, so that it will be possible to use them latter to extend the ABI.

The same reasonning apply to comp_mask field present in newer uverbs command: per commit 22878dbc9173, unsupported values in comp_mask are rejected.

[1] http://blog.ffwll.ch/2013/11/botching-up-ioctls.html

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs_main.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/infiniband/core/uverbs_main.c +3 -0

index 6c4fc6338b26..8652c13f6ea2 100644
@@ -669,6 +669,9 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count)
 			return -EINVAL;
 
+		if (ex_hdr.cmd_hdr_reserved)
+			return -EINVAL;
+
 		response = (char __user *)(unsigned long)ex_hdr.response;
 
 		if (response) {

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #7 lore raw

PATCH v4 6/10 IB/uverbs: check comp_mask in destroy_flow

In reply to a message not in the archive

Just like the check added to create_flow in 22878dbc9173, comp_mask must be checked in destroy_flow too.

Since only empty comp_mask is currently supported, any other value must be rejected.

This check was silently added in a previous patch[1] to move comp_mask in extended command header, part of previous patchset[2] against create/destroy_flow uverbs. The idea of moving comp_mask to the header was discarded for the final patchset[3].

Unfortunately the check added in destroy_flow uverb was not integrated in the final patchset.

[1] http://marc.info/?i=40175eda10d670d098204da6aa4c327a0171ae5f.1381510045.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org
[2] http://marc.info/?i=cover.1381510045.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org
[3] http://marc.info/?i=cover.1383773832.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Cc Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/[email protected]>

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs_cmd.c | 3 +++
 1 file changed, 3 insertions(+)

drivers/infiniband/core/uverbs_cmd.c +3 -0

index 971e16c970b9..a23874e959a1 100644
@@ -2833,6 +2833,9 @@ int ib_uverbs_ex_destroy_flow(struct ib_uverbs_file *file,
 	if (ret)
 		return ret;
 
+	if (cmd.comp_mask)
+		return -EINVAL;
+
 	uobj = idr_write_uobj(&ib_uverbs_rule_idr, cmd.flow_handle,
 			      file->ucontext);
 	if (!uobj)

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #8 lore raw

PATCH v4 7/10 IB/uverbs: check reserved fields in create_flow

In reply to a message not in the archive

As noted by Daniel Vetter in its article "Botching up ioctls"[1]

It's important to ensure that reserved fields are set to known value, so that it will be possible to use them latter to extend the ABI.

The same reasonning apply to comp_mask field present in newer uverbs command: per commit 22878dbc9173, unsupported values in comp_mask are rejected.

[1] http://blog.ffwll.ch/2013/11/botching-up-ioctls.html

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs_cmd.c | 7 +++++++
 1 file changed, 7 insertions(+)

drivers/infiniband/core/uverbs_cmd.c +7 -0

index a23874e959a1..f50f1f3dc509 100644
@@ -2631,6 +2631,9 @@ out_put:
 static int kern_spec_to_ib_spec(struct ib_uverbs_flow_spec *kern_spec,
 				union ib_flow_spec *ib_spec)
 {
+	if (kern_spec->reserved)
+		return -EINVAL;
+
 	ib_spec->type = kern_spec->type;
 
 	switch (ib_spec->type) {

@@ -2709,6 +2712,10 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	    (cmd.flow_attr.num_of_specs * sizeof(struct ib_uverbs_flow_spec)))
 		return -EINVAL;
 
+	if (cmd.flow_attr.reserved[0] ||
+	    cmd.flow_attr.reserved[1])
+		return -EINVAL;
+
 	if (cmd.flow_attr.num_of_specs) {
 		kern_flow_attr = kmalloc(sizeof(*kern_flow_attr) + cmd.flow_attr.size,
 					 GFP_KERNEL);

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #9 lore raw

PATCH v4 8/10 IB/uverbs: set error code when fail to consume all flow_spec items

In reply to a message not in the archive

If the flow_spec items parsed count does not match the number of items declared in the flow_attr command, or if not all bytes are used for flow_spec items (eg. trailing garbage), a log message is reported and the function leave through the error path. Unfortunately the error code is currently not set.

This patch set error code to -EINVAL in such cases, so that the error is reported to userspace instead of silently fail.

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

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

drivers/infiniband/core/uverbs_cmd.c +1 -0

index f50f1f3dc509..1288a7ce254a 100644
@@ -2776,6 +2776,7 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	if (cmd.flow_attr.size || (i != flow_attr->num_of_specs)) {
 		pr_warn("create flow failed, flow %d: %d bytes left from uverb cmd\n",
 			i, cmd.flow_attr.size);
+		err = -EINVAL;
 		goto err_free;
 	}
 	flow_id = ib_create_flow(qp, flow_attr, IB_FLOW_DOMAIN_USER);

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #10 lore raw

PATCH v4 9/10 IB/uverbs: check access to userspace response buffer in extended command

In reply to a message not in the archive

This patch adds a check on the output buffer with access_ok(VERIFY_WRITE, ...) to ensure the whole buffer is in userspace memory before using the pointer in uverbs functions. If the buffer or a subset of it is not valid, returns -EFAULT to the caller.

This will also catch invalid buffer before the final call to copy_to_user() which happen late in most uverb functions.

Just like the check in read(2) syscall, it's a sanity check to detect invalid parameters provided by userspace. This particular check was added in vfs_read() by Linus Torvalds for v2.6.12 with following commit message:

https://git.kernel.org/cgit/linux/kernel/git/tglx/history.git/commit/?id=fd770e66c9a65b14ce114e171266cf6f393df502

Make read/write always do the full "access_ok()" tests.

The actual user copy will do them too, but only for the range that ends up being actually copied. That hides bugs when the range has been clamped by file size or other issues.

Note: there's no need to check input buffer since vfs_write() already does access_ok(VERIFY_READ, ...) as part of write() syscall.

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs_main.c | 5 +++++
 1 file changed, 5 insertions(+)

drivers/infiniband/core/uverbs_main.c +5 -0

index 8652c13f6ea2..0be1dd86f768 100644
@@ -677,6 +677,11 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf,
 		if (response) {
 			if (!hdr.out_words && !ex_hdr.provider_out_words)
 				return -EINVAL;
+
+			if (!access_ok(VERIFY_WRITE,
+				       response,
+				       (hdr.out_words + ex_hdr.provider_out_words) * 8))
+				return -EFAULT;
 		} else {
 			if (hdr.out_words || ex_hdr.provider_out_words)
 				return -EINVAL;

Signature · 6 lines

1.8.4.2

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

Yann Droneaud

author

2013-12-17 09:58 UTC #11 lore raw

PATCH v4 10/10 IB/uverbs: check input length in flow steering uverbs

In reply to a message not in the archive

Since ib_copy_from_udata() doesn't check yet the available input data length before accessing userspace memory, an explicit check of this length is required to prevent:

- reading past the user provided buffer,
- underflow when subtracting the expected command size from the input length.

This will ensure the newly added flow steering uverbs don't try to process truncated commands.

Link http://marc.info/?i=cover.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Signed-off-by Yann Droneaud <[email protected]>

---
 drivers/infiniband/core/uverbs_cmd.c | 6 ++++++
 1 file changed, 6 insertions(+)

drivers/infiniband/core/uverbs_cmd.c +6 -0

index 1288a7ce254a..4a53d3f01dfc 100644
@@ -2687,6 +2687,9 @@ int ib_uverbs_ex_create_flow(struct ib_uverbs_file *file,
 	void *ib_spec;
 	int i;
 
+	if (ucore->inlen < sizeof(cmd))
+		return -EINVAL;
+
 	if (ucore->outlen < sizeof(resp))
 		return -ENOSPC;

@@ -2837,6 +2840,9 @@ int ib_uverbs_ex_destroy_flow(struct ib_uverbs_file *file,
 	struct ib_uobject		*uobj;
 	int				ret;
 
+	if (ucore->inlen < sizeof(cmd))
+		return -EINVAL;
+
 	ret = ib_copy_from_udata(&cmd, ucore, sizeof(cmd));
 	if (ret)
 		return ret;

Signature · 6 lines

1.8.4.2

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

Or Gerlitz

2013-12-17 10:35 UTC #12 lore raw

In reply to a message not in the archive

Quoted 4 lines · Yann Droneaud

Hi Roland,
Please find the fourth revision of a patchset against create_flow/destroy_flow and associated extended command scheme.

Yann, note that V3 is already applied, so you need to change incremental changes --
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): CAJZOPZKKg73s6Y+=KD6c8vdDXG0CmkL_gErv=KE-U0jPTSWEjg-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

Yann Droneaud

author

2013-12-19 13:21 UTC #14 lore raw

Hi Or, Roland.

Quoted 6 lines · Or Gerlitz

Quoted 3 lines · Yann Droneaud
Please find the fourth revision of a patchset against create_flow/destroy_flow and associated extended command scheme.
Yann, note that V3 is already applied, so you need to change incremental changes

Thanks for pointing this, Or.

Unfortunately, some patches in V4 were replacement for patches in V3. In this particular case, I've rework the commit messages in V4, so these changes might be lost.

In this other hand, V3 is not applied asis in for-next branch of Roland's tree. In fact Roland seems to have dropped this patch:
[PATCHv3 for-3.13 2/9] IB/uverbs: remove implicit cast in INIT_UDATA()
http://marc.info/?i=a6da021658e9a1cee5cf9f6db61694711cc2852d.1386798254.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

This, then, require a conflict fix when applying next patch:
[PATCHv3 for-3.13 3/9] IB/uverbs: set outbuf to NULL when no core response space is provided
http://marc.info/?i=d4ae6494fb2dc1f8a81e8d8e90d8084b37d15ba2.1386798254.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

And another for this later patch:
[PATCHv3 for-3.13 8/9] IB/uverbs: check access to userspace response
buffer in extended command
http://marc.info/?i=701fc3ee1136bbb4a0fd3d560c3ec1ee3bb3b828.1386798254.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

And I'm a bit disappointed about this particular patch applied with a conflict fix:

https://git.kernel.org/cgit/linux/kernel/git/roland/infiniband.git/commit/?h=for-next&id=2f92baf8313756fd32da4d2a24abc67c8f4f7026

	if (!access_ok(VERIFY_WRITE,
		       (const void __user *) (unsigned long) ex_hdr.response,
		       (hdr.out_words + ex_hdr.provider_out_words) * 8))
		return -EFAULT;

https://git.kernel.org/cgit/linux/kernel/git/roland/infiniband.git/tree/drivers/infiniband/core/uverbs_main.c?h=for-next&id=2f92baf8313756fd32da4d2a24abc67c8f4f7026#n679

Checking for write access on a pointer to const doesn't sound right. It's harmless, and, if sparse/coccinelle doesn't have a check for such, it should be added to report a warning.

BTW, I would prefer being responsible of my own mistakes :)

I try to be helpful and open to fix my patchset so that they are easier to merge. In this particular case, the fix to my commit was not 'agreed': I haven't the chance to review it before being commited since it wasn't discussed by mail.

So I hope the infiniband/for-next branch could be rewrote (it's a branch for linux-next, do people care of it being rebased ?) to include my new commit messages and discard the not so incorrect fix to access_ok() call. And perhaps include the patch adding a 'response' pointer as 'common pattern' of uverbs function.

Just tell me what you want me to do to ease the integration of my fixes.

Regards.

Signature · 8 lines

Yann Droneaud
OPTEYA


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

A message that is not in the archive (sent off-list or not yet received): 1387459287.11925.52.camel-bi+AKbBUZKY6gyzm1THtWbp2dZbC/[email protected]

Or Gerlitz

2013-12-19 13:25 UTC #16 lore raw

Quoted 1 line · Yann Droneaud

[...] So I hope the infiniband/for-next branch could be rewrote

I am OK with that.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

Roland Dreier

2013-12-19 17:08 UTC #17 lore raw

In reply to a message not in the archive

Quoted 14 lines

Unfortunately, some patches in V4 were replacement for patches in V3. In this particular case, I've rework the commit messages in V4, so these changes might be lost.
And I'm a bit disappointed about this particular patch applied with a conflict fix:
https://git.kernel.org/cgit/linux/kernel/git/roland/infiniband.git/commit/?h=for-next&id=2f92baf8313756fd32da4d2a24abc67c8f4f7026
        if (!access_ok(VERIFY_WRITE,
                       (const void __user *) (unsigned long) ex_hdr.response,
                       (hdr.out_words + ex_hdr.provider_out_words) * 8))
                return -EFAULT;

Quoted 5 lines

https://git.kernel.org/cgit/linux/kernel/git/roland/infiniband.git/tree/drivers/infiniband/core/uverbs_main.c?h=for-next&id=2f92baf8313756fd32da4d2a24abc67c8f4f7026#n679
Checking for write access on a pointer to const doesn't sound right. It's harmless, and, if sparse/coccinelle doesn't have a check for such, it should be added to report a warning.

Sorry for that mistake. I've fixed that conflict fix to get rid of the const. I've also pulled in the changelog updates.

However, I still prefer not moving all the casts out into every place where we use INIT_UDATA(), so I'd prefer something like the patch below. I integrated that into the series and pushed out the result (since everyone seems OK with rebasing for-next).

Please let me know what you think.

From: Roland Dreier <[email protected]>
Date: Thu, 19 Dec 2013 08:37:03 -0800
Subject: [PATCH] IB/uverbs: Set pointers to NULL if length is 0 in
INIT_UDATA()

Trying to have a ternary operator to choose between NULL (or 0) and the real pointer value in invocations leads to an impossible choice between a sparse error about a literal 0 used as a NULL pointer, and a gcc warning about "pointer/integer type mismatch in conditional expression."

Rather than clutter the source with more casts, move the ternary operator into the INIT_UDATA() macro, which makes it easier to use and simplifies its callers.

Reported-by Yann Droneaud <[email protected]>

Signed-off-by Roland Dreier <[email protected]>

---
 drivers/infiniband/core/uverbs.h      | 12 ++++++------
 drivers/infiniband/core/uverbs_main.c | 11 ++++-------
 2 files changed, 10 insertions(+), 13 deletions(-)

drivers/infiniband/core/uverbs.h +6 -6

index 9879568aed8c..d2defbbdb5da 100644
@@ -47,12 +47,12 @@
 #include <rdma/ib_umem.h>
 #include <rdma/ib_user_verbs.h>
 
-#define INIT_UDATA(udata, ibuf, obuf, ilen, olen)            \
-    do {                                \
-        (udata)->inbuf  = (const void __user *) (ibuf);        \
-        (udata)->outbuf = (void __user *) (obuf);        \
-        (udata)->inlen  = (ilen);                \
-        (udata)->outlen = (olen);                \
+#define INIT_UDATA(udata, ibuf, obuf, ilen, olen)                \
+    do {                                    \
+        (udata)->inbuf  = (ilen) ? (const void __user *) (ibuf) : NULL;    \
+        (udata)->outbuf = (olen) ? (void __user *) (obuf) : NULL;    \
+        (udata)->inlen  = (ilen);                    \
+        (udata)->outlen = (olen);                    \
     } while (0)
 
 /*

(quoted diff) +0 -0

b/drivers/infiniband/core/uverbs_main.c
index 34386943ebcf..76b6f217a13e 100644

drivers/infiniband/core/uverbs_main.c +0 -0

@@ -676,15 +676,12 @@ static ssize_t ib_uverbs_write(struct file

*filp, const char __user *buf,

                 return -EINVAL;
         }

- INIT_UDATA(&ucore,

-               (hdr.in_words) ? buf : 0,
-               (unsigned long)ex_hdr.response,
-               hdr.in_words * 8,
-               hdr.out_words * 8);

+ INIT_UDATA(&ucore, buf, (unsigned long)ex_hdr.response,
+ hdr.in_words * 8, hdr.out_words * 8);

INIT_UDATA(&uhw,
- (ex_hdr.provider_in_words) ? buf + ucore.inlen : 0,
- (ex_hdr.provider_out_words) ? (unsigned
long)ex_hdr.response + ucore.outlen : 0,
+ buf + ucore.inlen,
+ (unsigned long) ex_hdr.response + ucore.outlen,
ex_hdr.provider_in_words * 8, ex_hdr.provider_out_words * 8);
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): CAL1RGDUfC38aKqDv7rOnbi3m7PMPC8ze+1kymgrWwNvYHU7e2Q-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org

Yann Droneaud

author

2013-12-19 18:29 UTC #19 lore raw

Hi Roland,

Quoted 23 lines · Roland Dreier

Quoted 14 lines
Unfortunately, some patches in V4 were replacement for patches in V3. In this particular case, I've rework the commit messages in V4, so these changes might be lost.
And I'm a bit disappointed about this particular patch applied with a conflict fix:
https://git.kernel.org/cgit/linux/kernel/git/roland/infiniband.git/commit/?h=for-next&id=2f92baf8313756fd32da4d2a24abc67c8f4f7026
        if (!access_ok(VERIFY_WRITE,
                       (const void __user *) (unsigned long) ex_hdr.response,
                       (hdr.out_words + ex_hdr.provider_out_words) * 8))
                return -EFAULT;
Quoted 5 lines
https://git.kernel.org/cgit/linux/kernel/git/roland/infiniband.git/tree/drivers/infiniband/core/uverbs_main.c?h=for-next&id=2f92baf8313756fd32da4d2a24abc67c8f4f7026#n679
Checking for write access on a pointer to const doesn't sound right. It's harmless, and, if sparse/coccinelle doesn't have a check for such, it should be added to report a warning.
Sorry for that mistake. I've fixed that conflict fix to get rid of the const. I've also pulled in the changelog updates.

Quoted 7 lines

However, I still prefer not moving all the casts out into every place where we use INIT_UDATA(), so I'd prefer something like the patch below. I integrated that into the series and pushed out the result (since everyone seems OK with rebasing for-next).
Please let me know what you think.

I'm a bit puzzled :) Sure it address the warning problem from sparse and gcc. It's even handling setting NULL for response if out_words is 0, which was part of another patch[1] in the patchset. So I would say OK.

But I'm concerned about this global change of behavior of INIT_UDATA().

I felt it was OK to handle the case of out_words equal to 0 in the new uverbs, since there's no current user of this, there's nothing to broke. It was a limited, locally change.

But when changing globally the behavor of INIT_UDATA(), what about existing program which pass a non-NULL address for the response and set out_words to 0 in a command.

libibverbs doesn't seems to do this, but who knows how one might try to access to uverbs ?

Since currently ib_copy_to_udata() doesn't check struct ib_udata outlen (I have patch for that[2]), some provider (hw/) might be using ib_copy_to_udata() without paying attention to outlen.

If outbuf is set to NULL when outlen is 0, then it might broke, and this could be called a regression.

I haven't double check each and every user of ib_copy_to_udata(), even if I've found some ugly hack regarding udata, I'm looking at you ipath[3] and qib[4], I haven't payed yet attention to user of ib_copy_to_udata() without check for outlen first.

In this other hand, with my patch[2] from the patchset[5], I've done some tests, and haven't found any regression, but my testbed was very limited.

So I think we must double check here before applying a global change to INIT_UDATA() and/or be ready to pay the price of regression in some vendor software (I'm not aware of) ...

[1] [PATCHv4 for-3.13 04/10] IB/uverbs: set outbuf to NULL when no core response space is provided
http://marc.info/?i=330b13a0884b6cc03e287a32f8f49c1aac6bdbed.1387273677.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

[2] [PATCH 04/22] infiniband: ib_copy_to_udata(): check output length
http://marc.info/?i=d27716a3a1c180f832d153a7402f65ea8a75b734.1376847403.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

[3] ipath_srq.c:254 ipath_modify_srq()
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/infiniband/hw/ipath/ipath_srq.c?id=v3.13-rc4#n254

[4] qib_srq.c:250 qib_modify_srq()
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/infiniband/hw/qib/qib_srq.c?id=v3.13-rc4#n250

[5] [PATCH 00/22] infiniband: improve userspace input check
http://marc.info/?i=cover.1376847403.git.ydroneaud-RlY5vtjFyJ3QT0dZR+AlfA@public.gmane.org

Regards.

Signature · 8 lines

Yann Droneaud
OPTEYA


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

A message that is not in the archive (sent off-list or not yet received): 1387477777.11925.85.camel-bi+AKbBUZKY6gyzm1THtWbp2dZbC/[email protected]

Yann Droneaud

author

2013-12-19 22:57 UTC #21 lore raw

Hi,

Quoted 31 lines · Yann Droneaud

Quoted 8 lines · Roland Dreier
However, I still prefer not moving all the casts out into every place where we use INIT_UDATA(), so I'd prefer something like the patch below. I integrated that into the series and pushed out the result (since everyone seems OK with rebasing for-next).
Please let me know what you think.
I'm a bit puzzled :) Sure it address the warning problem from sparse and gcc. It's even handling setting NULL for response if out_words is 0, which was part of another patch[1] in the patchset. So I would say OK.
But I'm concerned about this global change of behavior of INIT_UDATA().
I felt it was OK to handle the case of out_words equal to 0 in the new uverbs, since there's no current user of this, there's nothing to broke. It was a limited, locally change.
But when changing globally the behavor of INIT_UDATA(), what about existing program which pass a non-NULL address for the response and set out_words to 0 in a command.
libibverbs doesn't seems to do this, but who knows how one might try to access to uverbs ?
Since currently ib_copy_to_udata() doesn't check struct ib_udata outlen (I have patch for that[2]), some provider (hw/) might be using ib_copy_to_udata() without paying attention to outlen.

Quoted 8 lines

If outbuf is set to NULL when outlen is 0, then it might broke, and this could be called a regression.
I haven't double check each and every user of ib_copy_to_udata(), even if I've found some ugly hack regarding udata, I'm looking at you ipath[3] and qib[4], I haven't payed yet attention to user of ib_copy_to_udata() without check for outlen first.

I took some time to read again the source code.

I've asked about the consequences for users of ib_copy_to_udata(), but ib_copy_from_udata() is also concerned by this issue, but very unlikely.

To demonstrate how it is possible to have valid (!) uverbs calls building a struct ib_udata with either outlen set 0 or inlen set 0, let's define some 'vendor' command and response structure in userspace:

    #include <infiniband/kern-abi.h>

    struct ibv_cmd_hdr {
        __u32 command;
        __u16 in_words;
        __u16 out_words;
    };

    struct vendor_create_cq {
        struct ibv_create_cq ibv_cmd;
        /* vendor defined fields */	
    };
 
    struct vendor_create_cq_resp resp {
        struct ibv_create_cq_resp ibv_resp;
        /* vendor defined fields */	
    };

Now, let's build a request through a custom 'libibverbs/libvendor-rdma' which might be valid under current kernel, but will set outlen to 0 while response/outbuf would be != NULL:

    struct ibv_cq *vendor_create_cq_broken_outlen(...)
    {
        struct vendor_create_cq      cmd;
        struct vendor_create_cq_resp resp;
        ...
        IBV_INIT_CMD_RESP(&cmd,
                          sizeof(cmd),
                          CREATE_CQ,
                          &resp,
                          sizeof(struct ibv_create_cq_resp)); /* <-- */
        ...
        write(context->cmd_fd,
              &cmd,
              sizeof(cmd));
        ...
    }

Now let's build a request which might be valid with current kernel but will set inlen to 0 while it will be possible for vendor driver (provider, under hw/) to read vendor fields:

    struct ibv_cq *vendor_create_cq_broken_inlen(...)
    {
        struct vendor_create_cq      cmd;
        struct vendor_create_cq_resp resp;
        size_t cmd_size = sizeof(struct ibv_create_cq) - 
                          sizeof(struct ibv_cmd_hdr); /* <-- */
        ...
        assert(cmd_size >= sizeof(struct ibv_cmd_hdr));
        ...
        IBV_INIT_CMD_RESP(&cmd,
                          cmd_size,
                          CREATE_CQ,
                          &resp,
                          sizeof(resp));
        ...
        write(context->cmd_fd,
              &cmd,
              cmd_size);
        ...
    }

It's harder to get it wrong here, because struct ib_udata inlen, for current uverbs (eg. not the extended one, currently only flow steering related), have not subtracted the command header length, while the inbuf pointer is added the command header length. This is a bit schizophrenic. It's making bound checking a little bit harder, see for example mthca_reg_user_mr(), which is the only function doing it right.
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/infiniband/hw/mthca/mthca_provider.c?id=v3.13-rc4#n988

(I also have a patch to subtract the size of header from input length,
so that my patches[1] to add proper bound checking become effective at checking the bounds).

So in order to trigger INIT_UDATA() to set inbuf to NULL, the total size of the command must be equal to the size of command, leaving apart the size of the header, while still required to be greater than the header to be accepted by ib_uverbs_write().

We can safely state that it is very unlikely that any unknown userspace program is doing this. But keep it mind it's possible.

So if INIT_UDATA() is changed to set inbuf to NULL if inlen is equal to 0, and to set outbuf to NULL if outlen is equal to 0, then, code using, respectively, ib_copy_from_udata() and ib_copy_to_udata() without checking the size will break for the previous code snippet.

I believe these are corner cases, and, in french, I would say it's "tiré par les cheveux" [2], but when looking at infiniband/hw/, we can see several (hint, *a lot of*) places where bound checking is not done.

(That's why I've suggested adding checks in ib_copy_{from,to}_udata(),
in my patchset[1][3][4], but it require subtracting the header length, fix mthca to not do it, fix ipath and qib to not swap the pointer behind the scene, and a lot of testing/reviewing to catch possible regression, but sometime these a desirable, but I digress).

Here's a list of calls to ib_copy_{from,to}_data() for which there's no or incorrect checking of {input,output} length:

- drivers/infiniband/hw/cxgb3/iwch_provider.c:162, iwch_create_cq()

ib_copy_from_udata() with no check on input length.

- drivers/infiniband/hw/cxgb3/iwch_provider.c:232, iwch_create_cq()

ib_copy_to_udata(), with a check on output length,
but will copy even if output length is 0.

- drivers/infiniband/hw/cxgb3/iwch_provider.c:436, iwch_allocate_pd()