2019-02-19 02:38:20

by Jann Horn

Subject: [PATCH] firmware: xilinx: fix debugfs write handler

- Userspace wants to write a string with `len` bytes, not counting the
terminating NULL, so we should allocate `len+1` bytes. It looks like the
current code relied on having a nullbyte directly behind `kern_buff`,
which happens to work reliably as long as `len` isn't one of the kmalloc
size classes.
- strncpy_from_user() is completely wrong here; userspace is giving us a
(not necessarily null-terminated) buffer and its length.
strncpy_from_user() is for cases in which we don't know the length.
- Don't let broken userspace allocate arbitrarily big kmalloc allocations.

Just use memdup_user_nul(), which is designed precisely for things like
this.

Signed-off-by: Jann Horn <[email protected]>
---
WARNING: completely untested patch

drivers/firmware/xilinx/zynqmp-debug.c | 15 ++++-----------
1 file changed, 4 insertions(+), 11 deletions(-)

diff --git a/drivers/firmware/xilinx/zynqmp-debug.c b/drivers/firmware/xilinx/zynqmp-debug.c
index 2771df6df379..90b66cdbfd58 100644
--- a/drivers/firmware/xilinx/zynqmp-debug.c
+++ b/drivers/firmware/xilinx/zynqmp-debug.c
@@ -163,21 +163,14 @@ static ssize_t zynqmp_pm_debugfs_api_write(struct file *file,

strcpy(debugfs_buf, "");

- if (*off != 0 || len == 0)
+ if (*off != 0 || len <= 1 || len > PAGE_SIZE - 1)
return -EINVAL;

- kern_buff = kzalloc(len, GFP_KERNEL);
- if (!kern_buff)
- return -ENOMEM;
-
+ kern_buff = memdup_user_nul(ptr, len);
+ if (IS_ERR(kern_buff))
+ return PTR_ERR(kern_buff);
tmp_buff = kern_buff;

- ret = strncpy_from_user(kern_buff, ptr, len);
- if (ret < 0) {
- ret = -EFAULT;
- goto err;
- }
-
/* Read the API name from a user request */
pm_api_req = strsep(&kern_buff, " ");

--
2.21.0.rc0.258.g878e2cd30e-goog

2019-02-28 14:00:11

by Michal Simek

Subject: Re: [PATCH] firmware: xilinx: fix debugfs write handler

On 18. 02. 19 22:43, Jann Horn wrote:
> - Userspace wants to write a string with `len` bytes, not counting the
> terminating NULL, so we should allocate `len+1` bytes. It looks like the
> current code relied on having a nullbyte directly behind `kern_buff`,
> which happens to work reliably as long as `len` isn't one of the kmalloc
> size classes.
> - strncpy_from_user() is completely wrong here; userspace is giving us a
> (not necessarily null-terminated) buffer and its length.
> strncpy_from_user() is for cases in which we don't know the length.
> - Don't let broken userspace allocate arbitrarily big kmalloc allocations.
>
> Just use memdup_user_nul(), which is designed precisely for things like
> this.
>
> Signed-off-by: Jann Horn <[email protected]>
> ---
> WARNING: completely untested patch
>
> drivers/firmware/xilinx/zynqmp-debug.c | 15 ++++-----------
> 1 file changed, 4 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/firmware/xilinx/zynqmp-debug.c b/drivers/firmware/xilinx/zynqmp-debug.c
> index 2771df6df379..90b66cdbfd58 100644
> --- a/drivers/firmware/xilinx/zynqmp-debug.c
> +++ b/drivers/firmware/xilinx/zynqmp-debug.c
> @@ -163,21 +163,14 @@ static ssize_t zynqmp_pm_debugfs_api_write(struct file *file,
>
> strcpy(debugfs_buf, "");
>
> - if (*off != 0 || len == 0)
> + if (*off != 0 || len <= 1 || len > PAGE_SIZE - 1)
> return -EINVAL;
>
> - kern_buff = kzalloc(len, GFP_KERNEL);
> - if (!kern_buff)
> - return -ENOMEM;
> -
> + kern_buff = memdup_user_nul(ptr, len);
> + if (IS_ERR(kern_buff))
> + return PTR_ERR(kern_buff);
> tmp_buff = kern_buff;
>
> - ret = strncpy_from_user(kern_buff, ptr, len);
> - if (ret < 0) {
> - ret = -EFAULT;
> - goto err;
> - }
> -
> /* Read the API name from a user request */
> pm_api_req = strsep(&kern_buff, " ");
>
>

Jolly: Can you please retest it and ACK?

Thanks,
Michal

2019-03-05 19:14:14

by Jolly Shah

Subject: RE: [PATCH] firmware: xilinx: fix debugfs write handler

Acked-by: Jolly Shah <[email protected]>


> -----Original Message-----
> From: Jann Horn <[email protected]>
> Sent: Monday, February 18, 2019 1:43 PM
> To: Michal Simek <[email protected]>; [email protected]
> Cc: Rajan Vaja <[email protected]>; Jolly Shah <[email protected]>; linux-
> [email protected]; [email protected]
> Subject: [PATCH] firmware: xilinx: fix debugfs write handler
>
> - Userspace wants to write a string with `len` bytes, not counting the
> terminating NULL, so we should allocate `len+1` bytes. It looks like the
> current code relied on having a nullbyte directly behind `kern_buff`,
> which happens to work reliably as long as `len` isn't one of the kmalloc
> size classes.
> - strncpy_from_user() is completely wrong here; userspace is giving us a
> (not necessarily null-terminated) buffer and its length.
> strncpy_from_user() is for cases in which we don't know the length.
> - Don't let broken userspace allocate arbitrarily big kmalloc allocations.
>
> Just use memdup_user_nul(), which is designed precisely for things like
> this.
>
> Signed-off-by: Jann Horn <[email protected]>
> ---
> WARNING: completely untested patch
>
> drivers/firmware/xilinx/zynqmp-debug.c | 15 ++++-----------
> 1 file changed, 4 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/firmware/xilinx/zynqmp-debug.c
> b/drivers/firmware/xilinx/zynqmp-debug.c
> index 2771df6df379..90b66cdbfd58 100644
> --- a/drivers/firmware/xilinx/zynqmp-debug.c
> +++ b/drivers/firmware/xilinx/zynqmp-debug.c
> @@ -163,21 +163,14 @@ static ssize_t zynqmp_pm_debugfs_api_write(struct
> file *file,
>
> strcpy(debugfs_buf, "");
>
> - if (*off != 0 || len == 0)
> + if (*off != 0 || len <= 1 || len > PAGE_SIZE - 1)
> return -EINVAL;
>
> - kern_buff = kzalloc(len, GFP_KERNEL);
> - if (!kern_buff)
> - return -ENOMEM;
> -
> + kern_buff = memdup_user_nul(ptr, len);
> + if (IS_ERR(kern_buff))
> + return PTR_ERR(kern_buff);
> tmp_buff = kern_buff;
>
> - ret = strncpy_from_user(kern_buff, ptr, len);
> - if (ret < 0) {
> - ret = -EFAULT;
> - goto err;
> - }
> -
> /* Read the API name from a user request */
> pm_api_req = strsep(&kern_buff, " ");
>
> --
> 2.21.0.rc0.258.g878e2cd30e-goog

2019-03-18 12:47:29

by Michal Simek

Subject: Re: [PATCH] firmware: xilinx: fix debugfs write handler

On 18. 02. 19 22:43, Jann Horn wrote:
> - Userspace wants to write a string with `len` bytes, not counting the
> terminating NULL, so we should allocate `len+1` bytes. It looks like the
> current code relied on having a nullbyte directly behind `kern_buff`,
> which happens to work reliably as long as `len` isn't one of the kmalloc
> size classes.
> - strncpy_from_user() is completely wrong here; userspace is giving us a
> (not necessarily null-terminated) buffer and its length.
> strncpy_from_user() is for cases in which we don't know the length.
> - Don't let broken userspace allocate arbitrarily big kmalloc allocations.
>
> Just use memdup_user_nul(), which is designed precisely for things like
> this.
>
> Signed-off-by: Jann Horn <[email protected]>
> ---
> WARNING: completely untested patch
>
> drivers/firmware/xilinx/zynqmp-debug.c | 15 ++++-----------
> 1 file changed, 4 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/firmware/xilinx/zynqmp-debug.c b/drivers/firmware/xilinx/zynqmp-debug.c
> index 2771df6df379..90b66cdbfd58 100644
> --- a/drivers/firmware/xilinx/zynqmp-debug.c
> +++ b/drivers/firmware/xilinx/zynqmp-debug.c
> @@ -163,21 +163,14 @@ static ssize_t zynqmp_pm_debugfs_api_write(struct file *file,
>
> strcpy(debugfs_buf, "");
>
> - if (*off != 0 || len == 0)
> + if (*off != 0 || len <= 1 || len > PAGE_SIZE - 1)
> return -EINVAL;
>
> - kern_buff = kzalloc(len, GFP_KERNEL);
> - if (!kern_buff)
> - return -ENOMEM;
> -
> + kern_buff = memdup_user_nul(ptr, len);
> + if (IS_ERR(kern_buff))
> + return PTR_ERR(kern_buff);
> tmp_buff = kern_buff;
>
> - ret = strncpy_from_user(kern_buff, ptr, len);
> - if (ret < 0) {
> - ret = -EFAULT;
> - goto err;
> - }
> -
> /* Read the API name from a user request */
> pm_api_req = strsep(&kern_buff, " ");
>
>

applied with Jolly's ack.

Thanks,
Michal