Skip to contentSkip to conversation

RDMA/rxe: Fix null deref on srq->rq.queue after resize failure

PATCH Superseded Zhu Yanjun 2 participants 4 messages Oct 27, 2025

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Zhu Yanjun

author
#0 lore raw

PATCH 1/1 RDMA/rxe: Fix null deref on srq->rq.queue after resize failure

A NULL pointer dereference can occur in rxe_srq_chk_attr() when ibv_modify_srq() is invoked twice in succession under certain error conditions. The first call may fail in rxe_queue_resize(), which leads rxe_srq_from_attr() to set srq->rq.queue = NULL. The second call then triggers a crash (null deref) when accessing srq->rq.queue->buf->index_mask.

Call Trace:
<TASK>
rxe_modify_srq+0x170/0x480 [rdma_rxe]
? __pfx_rxe_modify_srq+0x10/0x10 [rdma_rxe]
? uverbs_try_lock_object+0x4f/0xa0 [ib_uverbs]
? rdma_lookup_get_uobject+0x1f0/0x380 [ib_uverbs]
ib_uverbs_modify_srq+0x204/0x290 [ib_uverbs]
? __pfx_ib_uverbs_modify_srq+0x10/0x10 [ib_uverbs]
? tryinc_node_nr_active+0xe6/0x150
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x2c0/0x470 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_run_method+0x55a/0x6e0 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
ib_uverbs_cmd_verbs+0x54d/0x800 [ib_uverbs]
? __pfx_ib_uverbs_cmd_verbs+0x10/0x10 [ib_uverbs]
? __pfx___raw_spin_lock_irqsave+0x10/0x10
? __pfx_do_vfs_ioctl+0x10/0x10
? ioctl_has_perm.constprop.0.isra.0+0x2c7/0x4c0
? __pfx_ioctl_has_perm.constprop.0.isra.0+0x10/0x10
ib_uverbs_ioctl+0x13e/0x220 [ib_uverbs]
? __pfx_ib_uverbs_ioctl+0x10/0x10 [ib_uverbs]
__x64_sys_ioctl+0x138/0x1c0
do_syscall_64+0x82/0x250
? fdget_pos+0x58/0x4c0
? ksys_write+0xf3/0x1c0
? __pfx_ksys_write+0x10/0x10
? do_syscall_64+0xc8/0x250
? __pfx_vm_mmap_pgoff+0x10/0x10
? fget+0x173/0x230
? fput+0x2a/0x80
? ksys_mmap_pgoff+0x224/0x4c0
? do_syscall_64+0xc8/0x250
? do_user_addr_fault+0x37b/0xfe0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
entry_SYSCALL_64_after_hwframe+0x76/0x7e

Root cause:
The queue is released when the first failure of rxe_cq_resize_queue.
Thus, when rxe_cq_resize_queue is called again, the above call trace will occur.

Fix:
Aligning the error handling path in rxe_srq_from_attr() with rxe_cq_resize_queue(), which also uses rxe_queue_resize(): do not nullify the queue when resize fails.

Fixes 8700e3e7c485 ("Soft RoCE driver")
---
 drivers/infiniband/sw/rxe/rxe_srq.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
drivers/infiniband/sw/rxe/rxe_srq.c +1 -6
index 3661cb627d28..2a234f26ac10 100644
@@ -171,7 +171,7 @@ int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
 				       udata, mi, &srq->rq.producer_lock,
 				       &srq->rq.consumer_lock);
 		if (err)
-			goto err_free;
+			return err;
 
 		srq->rq.max_wr = attr->max_wr;
 	}
@@ -180,11 +180,6 @@ int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
 		srq->limit = attr->srq_limit;
 
 	return 0;
-
-err_free:
-	rxe_queue_cleanup(q);
-	srq->rq.queue = NULL;
-	return err;
 }
 
 void rxe_srq_cleanup(struct rxe_pool_elem *elem)
Signature · 1 line
2.51.0
author
#1 lore raw
Quoted 52 lines · Zhu Yanjun

A NULL pointer dereference can occur in rxe_srq_chk_attr() when ibv_modify_srq() is invoked twice in succession under certain error conditions. The first call may fail in rxe_queue_resize(), which leads rxe_srq_from_attr() to set srq->rq.queue = NULL. The second call then triggers a crash (null deref) when accessing srq->rq.queue->buf->index_mask.

Call Trace:
<TASK>
rxe_modify_srq+0x170/0x480 [rdma_rxe]
? __pfx_rxe_modify_srq+0x10/0x10 [rdma_rxe]
? uverbs_try_lock_object+0x4f/0xa0 [ib_uverbs]
? rdma_lookup_get_uobject+0x1f0/0x380 [ib_uverbs]
ib_uverbs_modify_srq+0x204/0x290 [ib_uverbs]
? __pfx_ib_uverbs_modify_srq+0x10/0x10 [ib_uverbs]
? tryinc_node_nr_active+0xe6/0x150
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x2c0/0x470 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_run_method+0x55a/0x6e0 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
ib_uverbs_cmd_verbs+0x54d/0x800 [ib_uverbs]
? __pfx_ib_uverbs_cmd_verbs+0x10/0x10 [ib_uverbs]
? __pfx___raw_spin_lock_irqsave+0x10/0x10
? __pfx_do_vfs_ioctl+0x10/0x10
? ioctl_has_perm.constprop.0.isra.0+0x2c7/0x4c0
? __pfx_ioctl_has_perm.constprop.0.isra.0+0x10/0x10
ib_uverbs_ioctl+0x13e/0x220 [ib_uverbs]
? __pfx_ib_uverbs_ioctl+0x10/0x10 [ib_uverbs]
__x64_sys_ioctl+0x138/0x1c0
do_syscall_64+0x82/0x250
? fdget_pos+0x58/0x4c0
? ksys_write+0xf3/0x1c0
? __pfx_ksys_write+0x10/0x10
? do_syscall_64+0xc8/0x250
? __pfx_vm_mmap_pgoff+0x10/0x10
? fget+0x173/0x230
? fput+0x2a/0x80
? ksys_mmap_pgoff+0x224/0x4c0
? do_syscall_64+0xc8/0x250
? do_user_addr_fault+0x37b/0xfe0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
entry_SYSCALL_64_after_hwframe+0x76/0x7e

Root cause:
The queue is released when the first failure of rxe_cq_resize_queue.
Thus, when rxe_cq_resize_queue is called again, the above call trace will occur.

Quoted 4 lines

Fix:
Aligning the error handling path in rxe_srq_from_attr() with rxe_cq_resize_queue(), which also uses rxe_queue_resize(): do not nullify the queue when resize fails.

This commit is based on the repository https://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma.git for-next.

Yanjun.Zhu

Quoted 35 lines
Fixes 8700e3e7c485 ("Soft RoCE driver")
---
  drivers/infiniband/sw/rxe/rxe_srq.c | 7 +------
  1 file changed, 1 insertion(+), 6 deletions(-)
drivers/infiniband/sw/rxe/rxe_srq.c +1 -6
index 3661cb627d28..2a234f26ac10 100644
@@ -171,7 +171,7 @@ int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
  				       udata, mi, &srq->rq.producer_lock,
  				       &srq->rq.consumer_lock);
  		if (err)
-			goto err_free;
+			return err;
 
  		srq->rq.max_wr = attr->max_wr;
  	}
@@ -180,11 +180,6 @@ int rxe_srq_from_attr(struct rxe_dev *rxe, struct rxe_srq *srq,
  		srq->limit = attr->srq_limit;
 
  	return 0;
-
-err_free:
-	rxe_queue_cleanup(q);
-	srq->rq.queue = NULL;
-	return err;
  }
 
  void rxe_srq_cleanup(struct rxe_pool_elem *elem)
#2 lore raw
In reply to Zhu Yanjun
Quoted 52 lines · Zhu Yanjun

A NULL pointer dereference can occur in rxe_srq_chk_attr() when ibv_modify_srq() is invoked twice in succession under certain error conditions. The first call may fail in rxe_queue_resize(), which leads rxe_srq_from_attr() to set srq->rq.queue = NULL. The second call then triggers a crash (null deref) when accessing srq->rq.queue->buf->index_mask.

Call Trace:
<TASK>
rxe_modify_srq+0x170/0x480 [rdma_rxe]
? __pfx_rxe_modify_srq+0x10/0x10 [rdma_rxe]
? uverbs_try_lock_object+0x4f/0xa0 [ib_uverbs]
? rdma_lookup_get_uobject+0x1f0/0x380 [ib_uverbs]
ib_uverbs_modify_srq+0x204/0x290 [ib_uverbs]
? __pfx_ib_uverbs_modify_srq+0x10/0x10 [ib_uverbs]
? tryinc_node_nr_active+0xe6/0x150
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x2c0/0x470 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_run_method+0x55a/0x6e0 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
ib_uverbs_cmd_verbs+0x54d/0x800 [ib_uverbs]
? __pfx_ib_uverbs_cmd_verbs+0x10/0x10 [ib_uverbs]
? __pfx___raw_spin_lock_irqsave+0x10/0x10
? __pfx_do_vfs_ioctl+0x10/0x10
? ioctl_has_perm.constprop.0.isra.0+0x2c7/0x4c0
? __pfx_ioctl_has_perm.constprop.0.isra.0+0x10/0x10
ib_uverbs_ioctl+0x13e/0x220 [ib_uverbs]
? __pfx_ib_uverbs_ioctl+0x10/0x10 [ib_uverbs]
__x64_sys_ioctl+0x138/0x1c0
do_syscall_64+0x82/0x250
? fdget_pos+0x58/0x4c0
? ksys_write+0xf3/0x1c0
? __pfx_ksys_write+0x10/0x10
? do_syscall_64+0xc8/0x250
? __pfx_vm_mmap_pgoff+0x10/0x10
? fget+0x173/0x230
? fput+0x2a/0x80
? ksys_mmap_pgoff+0x224/0x4c0
? do_syscall_64+0xc8/0x250
? do_user_addr_fault+0x37b/0xfe0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
entry_SYSCALL_64_after_hwframe+0x76/0x7e

Root cause:
The queue is released when the first failure of rxe_cq_resize_queue.
Thus, when rxe_cq_resize_queue is called again, the above call trace will occur.

Quoted 4 lines

Fix:
Aligning the error handling path in rxe_srq_from_attr() with rxe_cq_resize_queue(), which also uses rxe_queue_resize(): do not nullify the queue when resize fails.

Did these two paragraphs come from AI? They don't add any new information, let's remove them.

Quoted 3 lines

Link in "Closes" tag should point to report and not to some random place.

Quoted 6 lines
Fixes 8700e3e7c485 ("Soft RoCE driver")
---
 drivers/infiniband/sw/rxe/rxe_srq.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

It is second version of previously sent patch. Please add changelog.

Thanks

author
#3 lore raw
Quoted 68 lines · Leon Romanovsky
Quoted 52 lines · Zhu Yanjun

A NULL pointer dereference can occur in rxe_srq_chk_attr() when ibv_modify_srq() is invoked twice in succession under certain error conditions. The first call may fail in rxe_queue_resize(), which leads rxe_srq_from_attr() to set srq->rq.queue = NULL. The second call then triggers a crash (null deref) when accessing srq->rq.queue->buf->index_mask.

Call Trace:
<TASK>
rxe_modify_srq+0x170/0x480 [rdma_rxe]
? __pfx_rxe_modify_srq+0x10/0x10 [rdma_rxe]
? uverbs_try_lock_object+0x4f/0xa0 [ib_uverbs]
? rdma_lookup_get_uobject+0x1f0/0x380 [ib_uverbs]
ib_uverbs_modify_srq+0x204/0x290 [ib_uverbs]
? __pfx_ib_uverbs_modify_srq+0x10/0x10 [ib_uverbs]
? tryinc_node_nr_active+0xe6/0x150
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x2c0/0x470 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
? uverbs_fill_udata+0xed/0x4f0 [ib_uverbs]
ib_uverbs_run_method+0x55a/0x6e0 [ib_uverbs]
? __pfx_ib_uverbs_handler_UVERBS_METHOD_INVOKE_WRITE+0x10/0x10 [ib_uverbs]
ib_uverbs_cmd_verbs+0x54d/0x800 [ib_uverbs]
? __pfx_ib_uverbs_cmd_verbs+0x10/0x10 [ib_uverbs]
? __pfx___raw_spin_lock_irqsave+0x10/0x10
? __pfx_do_vfs_ioctl+0x10/0x10
? ioctl_has_perm.constprop.0.isra.0+0x2c7/0x4c0
? __pfx_ioctl_has_perm.constprop.0.isra.0+0x10/0x10
ib_uverbs_ioctl+0x13e/0x220 [ib_uverbs]
? __pfx_ib_uverbs_ioctl+0x10/0x10 [ib_uverbs]
__x64_sys_ioctl+0x138/0x1c0
do_syscall_64+0x82/0x250
? fdget_pos+0x58/0x4c0
? ksys_write+0xf3/0x1c0
? __pfx_ksys_write+0x10/0x10
? do_syscall_64+0xc8/0x250
? __pfx_vm_mmap_pgoff+0x10/0x10
? fget+0x173/0x230
? fput+0x2a/0x80
? ksys_mmap_pgoff+0x224/0x4c0
? do_syscall_64+0xc8/0x250
? do_user_addr_fault+0x37b/0xfe0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
? clear_bhb_loop+0x50/0xa0
entry_SYSCALL_64_after_hwframe+0x76/0x7e

Root cause:
The queue is released when the first failure of rxe_cq_resize_queue.
Thus, when rxe_cq_resize_queue is called again, the above call trace will occur.

Quoted 4 lines

Fix:
Aligning the error handling path in rxe_srq_from_attr() with rxe_cq_resize_queue(), which also uses rxe_queue_resize(): do not nullify the queue when resize fails.

Did these two paragraphs come from AI? They don't add any new information, let's remove them.

Quoted 3 lines

Link in "Closes" tag should point to report and not to some random place.

Quoted 8 lines
Quoted 6 lines
Fixes 8700e3e7c485 ("Soft RoCE driver")
---
  drivers/infiniband/sw/rxe/rxe_srq.c | 7 +------
  1 file changed, 1 insertion(+), 6 deletions(-)

It is second version of previously sent patch. Please add changelog.

OK. I will send the 3rd patch.

Yanjun.Zhu

Quoted 2 lines

Thanks