2016-02-25 22:29:05

by Colin King

[permalink] [raw]
Subject: [PATCH] mt7601u: do not free dma_buf when ivp allocation fails

From: Colin Ian King <[email protected]>

If the allocation of ivp fails the error handling attempts to
free an uninitialized dma_buf; this data structure just contains
garbage on the stack, so the freeing will cause issues when the
urb, buf and dma fields are cleaned. Fix this by handling the
ivp check and

Signed-off-by: Colin Ian King <[email protected]>
---
drivers/net/wireless/mediatek/mt7601u/mcu.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/wireless/mediatek/mt7601u/mcu.c b/drivers/net/wireless/mediatek/mt7601u/mcu.c
index fbb1986..06ac657 100644
--- a/drivers/net/wireless/mediatek/mt7601u/mcu.c
+++ b/drivers/net/wireless/mediatek/mt7601u/mcu.c
@@ -359,13 +359,13 @@ mt7601u_upload_firmware(struct mt7601u_dev *dev, const struct mt76_fw *fw)
struct mt7601u_dma_buf dma_buf;
void *ivb;
u32 ilm_len, dlm_len;
- int i, ret;
+ int i, ret = -ENOMEM;

ivb = kmemdup(fw->ivb, sizeof(fw->ivb), GFP_KERNEL);
- if (!ivb || mt7601u_usb_alloc_buf(dev, MCU_FW_URB_SIZE, &dma_buf)) {
- ret = -ENOMEM;
+ if (!ivb)
+ goto error_ivb;
+ if (mt7601u_usb_alloc_buf(dev, MCU_FW_URB_SIZE, &dma_buf))
goto error;
- }

ilm_len = le32_to_cpu(fw->hdr.ilm_len) - sizeof(fw->ivb);
dev_dbg(dev->dev, "loading FW - ILM %u + IVB %zu\n",
@@ -396,8 +396,9 @@ mt7601u_upload_firmware(struct mt7601u_dev *dev, const struct mt76_fw *fw)

dev_dbg(dev->dev, "Firmware running!\n");
error:
- kfree(ivb);
mt7601u_usb_free_buf(dev, &dma_buf);
+error_ivb:
+ kfree(ivb);

return ret;
}
--
2.7.0



2016-02-25 23:03:16

by Jakub Kicinski

[permalink] [raw]
Subject: Re: [PATCH] mt7601u: do not free dma_buf when ivp allocation fails

On 25 February 2016 17:28:59 GMT-05:00, Colin King <[email protected]> wrote:
>From: Colin Ian King <[email protected]>
>
>If the allocation of ivp fails the error handling attempts to
>free an uninitialized dma_buf; this data structure just contains
>garbage on the stack, so the freeing will cause issues when the
>urb, buf and dma fields are cleaned. Fix this by handling the
>ivp check and

Commit message looks truncated?

>Signed-off-by: Colin Ian King <[email protected]>
>---
> drivers/net/wireless/mediatek/mt7601u/mcu.c | 11 ++++++-----
> 1 file changed, 6 insertions(+), 5 deletions(-)
>
>diff --git a/drivers/net/wireless/mediatek/mt7601u/mcu.c
>b/drivers/net/wireless/mediatek/mt7601u/mcu.c
>index fbb1986..06ac657 100644
>--- a/drivers/net/wireless/mediatek/mt7601u/mcu.c
>+++ b/drivers/net/wireless/mediatek/mt7601u/mcu.c
>@@ -359,13 +359,13 @@ mt7601u_upload_firmware(struct mt7601u_dev *dev,
>const struct mt76_fw *fw)
> struct mt7601u_dma_buf dma_buf;
> void *ivb;
> u32 ilm_len, dlm_len;
>- int i, ret;
>+ int i, ret = -ENOMEM;

No need...

>
> ivb = kmemdup(fw->ivb, sizeof(fw->ivb), GFP_KERNEL);
>- if (!ivb || mt7601u_usb_alloc_buf(dev, MCU_FW_URB_SIZE, &dma_buf)) {
>- ret = -ENOMEM;
>+ if (!ivb)
>+ goto error_ivb;

...just return -ENOMEM here, because...

>+error_ivb:
>+ kfree(ivb);

...calling kfree on a null pointer is redundant.

Thanks!!
[trimming To/CC a bit]