mhi_alloc_controller() allocates a memory space for mhi_ctrl. When some
errors occur, mhi_ctrl should be freed by mhi_free_controller(). But
when ath11k_mhi_read_addr_from_dt() fails, the function returns without
calling mhi_free_controller(), which will lead to a memory leak.
We can fix it by calling mhi_free_controller() when
ath11k_mhi_read_addr_from_dt() fails.
Signed-off-by: Jianglei Nie <[email protected]>
---
drivers/net/wireless/ath/ath11k/mhi.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/net/wireless/ath/ath11k/mhi.c b/drivers/net/wireless/ath/ath11k/mhi.c
index fc3524e83e52..3318c7c2b32b 100644
--- a/drivers/net/wireless/ath/ath11k/mhi.c
+++ b/drivers/net/wireless/ath/ath11k/mhi.c
@@ -376,8 +376,10 @@ int ath11k_mhi_register(struct ath11k_pci *ab_pci)
if (test_bit(ATH11K_FLAG_FIXED_MEM_RGN, &ab->dev_flags)) {
ret = ath11k_mhi_read_addr_from_dt(mhi_ctrl);
- if (ret < 0)
+ if (ret < 0) {
+ mhi_free_controller(mhi_ctrl);
return ret;
+ }
} else {
mhi_ctrl->iova_start = 0;
mhi_ctrl->iova_stop = 0xFFFFFFFF;
--
2.25.1
On 5/26/2022 3:02 AM, Jianglei Nie wrote:
> mhi_alloc_controller() allocates a memory space for mhi_ctrl. When some
> errors occur, mhi_ctrl should be freed by mhi_free_controller(). But
> when ath11k_mhi_read_addr_from_dt() fails, the function returns without
> calling mhi_free_controller(), which will lead to a memory leak.
>
> We can fix it by calling mhi_free_controller() when
> ath11k_mhi_read_addr_from_dt() fails.
>
> Signed-off-by: Jianglei Nie <[email protected]>
Reviewed-by: Jeff Johnson <[email protected]>
> ---
> drivers/net/wireless/ath/ath11k/mhi.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/wireless/ath/ath11k/mhi.c b/drivers/net/wireless/ath/ath11k/mhi.c
> index fc3524e83e52..3318c7c2b32b 100644
> --- a/drivers/net/wireless/ath/ath11k/mhi.c
> +++ b/drivers/net/wireless/ath/ath11k/mhi.c
> @@ -376,8 +376,10 @@ int ath11k_mhi_register(struct ath11k_pci *ab_pci)
>
> if (test_bit(ATH11K_FLAG_FIXED_MEM_RGN, &ab->dev_flags)) {
> ret = ath11k_mhi_read_addr_from_dt(mhi_ctrl);
> - if (ret < 0)
> + if (ret < 0) {
> + mhi_free_controller(mhi_ctrl);
> return ret;
> + }
> } else {
> mhi_ctrl->iova_start = 0;
> mhi_ctrl->iova_stop = 0xFFFFFFFF;
This patch looks good to me. However there is an additional issue that
IMO should be addressed in this function with a separate patch.
Just after mhi_ctrl is allocated the pointer is saved:
ab_pci->mhi_ctrl = mhi_ctrl;
When there is an error, mhi_ctrl is freed, but ab_pci->mhi_ctrl still
has a dangling pointer to the freed memory. This has the potential to
result in a subsequent use-after-free or double-free error.
So IMO we should do one of two things (I haven't looked at the code in
detail to determine which is better):
1) don't set ab_pci->mhi_ctrl = mhi_ctrl until the end of the function
after success from mhi_register_controller(). This requires that none of
the functions called between the current assignment point and the
proposed assignment point dereference ab_pci->mhi_ctrl
2) set ab_pci->mhi_ctrl = NULL in all of the places where we call
mhi_free_controller()
2a) have a central error exit label and have all of the error conditions
goto that label where there is a central code:
mhi_free_controller(mhi_ctrl);
ab_pci->mhi_ctrl = NULL;
return ret;