Skip to contentSkip to conversation

IB/hfi1: Fix possible null-pointer dereference in _extend_sdma_tx_descs()

PATCH In review Tuo Li 3 participants 3 messages Aug 6, 2021 – Aug 19, 2021

Subsystems (2)
InfiniBand HFI1 driver
lore
Share & tools
lore

Tuo Li

author
#0 lore raw

PATCH v3 1/1 IB/hfi1: Fix possible null-pointer dereference in _extend_sdma_tx_descs()

kmalloc_array() is called to allocate memory for tx->descp. If it fails, the function __sdma_txclean() is called:
__sdma_txclean(dd, tx);

However, in the function __sdma_txclean(), tx-descp is dereferenced if tx->num_desc is not zero:
sdma_unmap_desc(dd, &tx->descp[0]);

To fix this possible null-pointer dereference, assign the return value of kmalloc_array() to a local variable descp, and then assign it to tx->descp if it is not NULL. Otherwise, go to enomem.

Fixes 7724105686e7 ("IB/hfi1: add driver files")
Reported-by TOTE Robot <[email protected]>
Signed-off-by Tuo Li <[email protected]>

---
v3:
* Add fixes line.
Thank Jason Gunthorpe for helpful advice.
v2:
* Assign the return value of kmalloc_array() to a local variable and then check it instead of assigning 0 to tx->num_desc when memory allocation fails.
Thank Mike Marciniszyn for helpful advice.

---
 drivers/infiniband/hw/hfi1/sdma.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)
drivers/infiniband/hw/hfi1/sdma.c +6 -5
index eb15c310d63d..6ca8d065f44e 100644
@@ -3055,6 +3055,7 @@ static void __sdma_process_event(struct sdma_engine *sde,
 static int _extend_sdma_tx_descs(struct hfi1_devdata *dd, struct sdma_txreq *tx)
 {
 	int i;
+	struct sdma_desc *descp;
 
 	/* Handle last descriptor */
 	if (unlikely((tx->num_desc == (MAX_DESC - 1)))) {

@@ -3075,12 +3076,12 @@ static int _extend_sdma_tx_descs(struct hfi1_devdata *dd, struct sdma_txreq *tx)
 	if (unlikely(tx->num_desc == MAX_DESC))
 		goto enomem;
 
-	tx->descp = kmalloc_array(
-			MAX_DESC,
-			sizeof(struct sdma_desc),
-			GFP_ATOMIC);
-	if (!tx->descp)
+	descp = kmalloc_array(MAX_DESC,
+					sizeof(struct sdma_desc),
+					GFP_ATOMIC);
+	if (!descp)
 		goto enomem;
+	tx->descp = descp;
 
 	/* reserve last descriptor for coalescing */
 	tx->desc_limit = MAX_DESC - 1;
Signature · 1 line
2.25.1
Tested-by Acked-by on v3 1/1
#1 lore raw
Quoted 11 lines
Fixes 7724105686e7 ("IB/hfi1: add driver files")
Reported-by TOTE Robot <[email protected]>
Signed-off-by Tuo Li <[email protected]>

---
v3:
* Add fixes line.
Thank Jason Gunthorpe for helpful advice.
v2:
* Assign the return value of kmalloc_array() to a local variable and then check it instead of assigning 0 to tx->num_desc when memory allocation fails. Thank Mike Marciniszyn for helpful advice.

Thanks!

I'm curious, did you find this with some fault injection testing?

Tested-by Mike Marciniszyn <[email protected]>
Acked-by Mike Marciniszyn <[email protected]>
maintainer
#2 lore raw
In reply to Tuo Li
Quoted 26 lines · Tuo Li

kmalloc_array() is called to allocate memory for tx->descp. If it fails, the function __sdma_txclean() is called:
__sdma_txclean(dd, tx);

However, in the function __sdma_txclean(), tx-descp is dereferenced if tx->num_desc is not zero:
sdma_unmap_desc(dd, &tx->descp[0]);

To fix this possible null-pointer dereference, assign the return value of kmalloc_array() to a local variable descp, and then assign it to tx->descp if it is not NULL. Otherwise, go to enomem.

Fixes 7724105686e7 ("IB/hfi1: add driver files")
Reported-by TOTE Robot <[email protected]>
Signed-off-by Tuo Li <[email protected]>
Tested-by Mike Marciniszyn <[email protected]>
Acked-by Mike Marciniszyn <[email protected]>

---
v3:
* Add fixes line.
Thank Jason Gunthorpe for helpful advice.
v2:
* Assign the return value of kmalloc_array() to a local variable and then check it instead of assigning 0 to tx->num_desc when memory allocation fails.
Thank Mike Marciniszyn for helpful advice.

Quoted 3 lines
---
 drivers/infiniband/hw/hfi1/sdma.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

I fixed the wonky code formatting and applied to for-rc, thanks

Jason