2023-10-13 08:27:16

by Alexis Lothoré

Subject: [PATCH] wifi: wilc1000: use vmm_table as array in wilc struct

From: Ajay Singh <[email protected]>

Enabling KASAN and running some iperf tests raises some memory issues with
vmm_table:

BUG: KASAN: slab-out-of-bounds in wilc_wlan_handle_txq+0x6ac/0xdb4
Write of size 4 at addr c3a61540 by task wlan0-tx/95

KASAN detects that we are writing data beyond range allocated to vmm_table.
There is indeed a mismatch between the size passed to allocator in
wilc_wlan_init, and the range of possible indexes used later: allocation
size is missing a multiplication by sizeof(u32)

While at it, instead of simply multiplying the allocation size, do not keep
dedicated dynamic allocation for vmm_table: define it as an array with the
relevant size in wilc struct, which is already dynamically allocated

Fixes: 40b717bfcefa ("wifi: wilc1000: fix DMA on stack objects")
Cc: [email protected]
Signed-off-by: Ajay Singh <[email protected]>
Signed-off-by: Alexis Lothoré <[email protected]>
---
drivers/net/wireless/microchip/wilc1000/netdev.h | 2 +-
drivers/net/wireless/microchip/wilc1000/wlan.c | 12 ------------
2 files changed, 1 insertion(+), 13 deletions(-)

diff --git a/drivers/net/wireless/microchip/wilc1000/netdev.h b/drivers/net/wireless/microchip/wilc1000/netdev.h
index bb1a315a7b7e..2137ef294953 100644
--- a/drivers/net/wireless/microchip/wilc1000/netdev.h
+++ b/drivers/net/wireless/microchip/wilc1000/netdev.h
@@ -245,7 +245,7 @@ struct wilc {
u8 *rx_buffer;
u32 rx_buffer_offset;
u8 *tx_buffer;
- u32 *vmm_table;
+ u32 vmm_table[WILC_VMM_TBL_SIZE];

struct txq_handle txq[NQUEUES];
int txq_entries;
diff --git a/drivers/net/wireless/microchip/wilc1000/wlan.c b/drivers/net/wireless/microchip/wilc1000/wlan.c
index 58bbf50081e4..d93493c40e49 100644
--- a/drivers/net/wireless/microchip/wilc1000/wlan.c
+++ b/drivers/net/wireless/microchip/wilc1000/wlan.c
@@ -1252,8 +1252,6 @@ void wilc_wlan_cleanup(struct net_device *dev)
while ((rqe = wilc_wlan_rxq_remove(wilc)))
kfree(rqe);

- kfree(wilc->vmm_table);
- wilc->vmm_table = NULL;
kfree(wilc->rx_buffer);
wilc->rx_buffer = NULL;
kfree(wilc->tx_buffer);
@@ -1491,14 +1489,6 @@ int wilc_wlan_init(struct net_device *dev)
goto fail;
}

- if (!wilc->vmm_table)
- wilc->vmm_table = kzalloc(WILC_VMM_TBL_SIZE, GFP_KERNEL);
-
- if (!wilc->vmm_table) {
- ret = -ENOBUFS;
- goto fail;
- }
-
if (!wilc->tx_buffer)
wilc->tx_buffer = kmalloc(WILC_TX_BUFF_SIZE, GFP_KERNEL);

@@ -1523,8 +1513,6 @@ int wilc_wlan_init(struct net_device *dev)
return 0;

fail:
- kfree(wilc->vmm_table);
- wilc->vmm_table = NULL;
kfree(wilc->rx_buffer);
wilc->rx_buffer = NULL;
kfree(wilc->tx_buffer);

---
base-commit: f28d2198de8cbefa17286d5182337a1d6d518643
change-id: 20231012-wilc1000_tx_oops-58ce91ee3e93

Best regards,
--
Alexis Lothoré, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

2023-10-13 09:24:42

by Michael Walle

Subject: Re: [PATCH] wifi: wilc1000: use vmm_table as array in wilc struct

Hi,

Am 2023-10-13 10:26, schrieb Alexis Lothoré:
> From: Ajay Singh <[email protected]>
>
> Enabling KASAN and running some iperf tests raises some memory issues
> with
> vmm_table:
>
> BUG: KASAN: slab-out-of-bounds in wilc_wlan_handle_txq+0x6ac/0xdb4
> Write of size 4 at addr c3a61540 by task wlan0-tx/95
>
> KASAN detects that we are writing data beyond range allocated to
> vmm_table.
> There is indeed a mismatch between the size passed to allocator in
> wilc_wlan_init, and the range of possible indexes used later:
> allocation
> size is missing a multiplication by sizeof(u32)

Nice catch.

> While at it, instead of simply multiplying the allocation size, do not
> keep
> dedicated dynamic allocation for vmm_table: define it as an array with
> the
> relevant size in wilc struct, which is already dynamically allocated
>
> Fixes: 40b717bfcefa ("wifi: wilc1000: fix DMA on stack objects")
> Cc: [email protected]
> Signed-off-by: Ajay Singh <[email protected]>
> Signed-off-by: Alexis Lothoré <[email protected]>

Looks good to me. But you'll change the alignment of the table, not sure
if that matters for some DMA controllers.

-michael

2023-10-16 06:58:31

by Alexis Lothoré

Subject: Re: [PATCH] wifi: wilc1000: use vmm_table as array in wilc struct

Hello Michael,

On 10/13/23 11:24, Michael Walle wrote:
>> Fixes: 40b717bfcefa ("wifi: wilc1000: fix DMA on stack objects")
>> Cc: [email protected]
>> Signed-off-by: Ajay Singh <[email protected]>
>> Signed-off-by: Alexis Lothoré <[email protected]>
>
> Looks good to me. But you'll change the alignment of the table, not sure
> if that matters for some DMA controllers.

Thanks for the review.
Indeed, I may have overlooked that point. I am not sure either how much of an
issue it could be, but checking back the driver history on lore, I see that it
has already been mentioned, so let's be safe and keep a dedicated kzalloc for
vmm_table to make sure its address is safe to use with DMA.
I will send the corresponding v2.

Thanks,
Alexis

>
> -michael

--
Alexis Lothoré, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com