RDMA/ocrdma: fix use after free in ocrdma_dealloc_ucontext_pd()
Subsystems (2)
PATCH 1/1 RDMA/ocrdma: fix use after free in ocrdma_dealloc_ucontext_pd()
From: Tom Rix <[email protected]>
In ocrdma_dealloc_ucontext_pd() uctx->cntxt_pd is assigned to the variable pd and then after uctx->cntxt_pd is freed, the variable pd is passed to function _ocrdma_dealloc_pd() which dereferences pd directly or through its call to ocrdma_mbx_dealloc_pd().
Reorder the free using the variable pd.
--- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/infiniband/hw/ocrdma/ocrdma_verbs.c +1 -1
@@ -434,9 +434,9 @@ static void ocrdma_dealloc_ucontext_pd(struct ocrdma_ucontext *uctx)pr_err ("%s(%d) Freeing in use pdid=0x%x.\n" , __func__, dev->id, pd->id); } -kfree (uctx->cntxt_pd); uctx->cntxt_pd =NULL ;_ocrdma_dealloc_pd (dev, pd); +kfree (pd); }static struct ocrdma_pd *ocrdma_get_ucontext_pd (struct ocrdma_ucontext *uctx)
Signature · 1 line
2.27.0
Quoted 8 lines · On Tue, Dec 29, 2020 at 06:46:53PM -0800, [email protected] wrote:
From: Tom Rix <[email protected]>
In
ocrdma_dealloc_ucontext_pd()uctx->cntxt_pdis assigned to the variable pd and then afteructx->cntxt_pdis freed, the variable pd is passed to function_ocrdma_dealloc_pd()which dereferences pd directly or through its call toocrdma_mbx_dealloc_pd().
Quoted 8 lines
Reorder the free using the variable pd.
Fixes 21a428a019c9 ("RDMA: Handle PD allocations by IB/core")Signed-off-by Tom Rix <[email protected]>--- drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Thanks,
Quoted 8 lines · On Tue, Dec 29, 2020 at 06:46:53PM -0800, [email protected] wrote:
From: Tom Rix <[email protected]>
In
ocrdma_dealloc_ucontext_pd()uctx->cntxt_pdis assigned to the variable pd and then afteructx->cntxt_pdis freed, the variable pd is passed to function_ocrdma_dealloc_pd()which dereferences pd directly or through its call toocrdma_mbx_dealloc_pd().
Quoted 6 lines
Reorder the free using the variable pd.
Fixes 21a428a019c9 ("RDMA: Handle PD allocations by IB/core")Signed-off-by Tom Rix <[email protected]>drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Applied to for-rc
Is anyone testing ocrdma? Just doing the pyverbs rdma tests with kasn turned on would have instantly caught this, and the change is nearly a year old.
Is ocrdma obsolete enough we can delete the driver?
Thanks,
Jason
Quoted 15 lines · Jason Gunthorpe
Quoted 8 lines · On Tue, Dec 29, 2020 at 06:46:53PM -0800, [email protected] wrote:
From: Tom Rix <[email protected]>
In
ocrdma_dealloc_ucontext_pd()uctx->cntxt_pdis assigned to the variable pd and then afteructx->cntxt_pdis freed, the variable pd is passed to function_ocrdma_dealloc_pd()which dereferences pd directly or through its call toocrdma_mbx_dealloc_pd().Quoted 6 lines
Reorder the free using the variable pd.
Fixes 21a428a019c9 ("RDMA: Handle PD allocations by IB/core")Signed-off-by Tom Rix <[email protected]>drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 7 lines
Applied to for-rc
Is anyone testing ocrdma? Just doing the pyverbs rdma tests with kasn turned on would have instantly caught this, and the change is nearly a year old.
Is ocrdma obsolete enough we can delete the driver?
I am not an authority on ocrdma, i am fixing treewide, the problems clang static analysis flags.
Tom
Quoted 4 lines
Thanks,
Jason
Selvin Xavier
maintainerQuoted 18 lines · Tom Rix
Quoted 15 lines · Jason Gunthorpe
Quoted 14 lines · On Tue, Dec 29, 2020 at 06:46:53PM -0800, [email protected] wrote:
From: Tom Rix <[email protected]>
In
ocrdma_dealloc_ucontext_pd()uctx->cntxt_pdis assigned to the variable pd and then afteructx->cntxt_pdis freed, the variable pd is passed to function_ocrdma_dealloc_pd()which dereferences pd directly or through its call toocrdma_mbx_dealloc_pd().Reorder the free using the variable pd.
Fixes 21a428a019c9 ("RDMA: Handle PD allocations by IB/core")Signed-off-by Tom Rix <[email protected]>drivers/infiniband/hw/ocrdma/ocrdma_verbs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 7 lines
Quoted 7 lines
Applied to for-rc
Is anyone testing ocrdma? Just doing the pyverbs rdma tests with kasn turned on would have instantly caught this, and the change is nearly a year old.
Is ocrdma obsolete enough we can delete the driver?
Broadcom is not doing any active development/testing with ocrdma now. I am checking with other teams to see if this can be deleted completely. Will get back asap.
Thanks,
Selvin
Quoted 10 lines
I am not an authority on ocrdma, i am fixing treewide, the problems clang static analysis flags.
Tom
Quoted 4 lines
Thanks,
Jason