We can use kvfree() instead of calling kfree() and vfree() based on
if-else and param_flags. kvfree() will check the type of address and
will call the respective function to free it.
Additionally we can also remove the use of DM_PARAMS_KMALLOC and
DM_PARAMS_VMALLOC.
Signed-off-by: Sudip Mukherjee <[email protected]>
---
drivers/md/dm-ioctl.c | 11 +----------
1 file changed, 1 insertion(+), 10 deletions(-)
diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
index 2adf81d..d5df3a5 100644
--- a/drivers/md/dm-ioctl.c
+++ b/drivers/md/dm-ioctl.c
@@ -1670,8 +1670,6 @@ static int check_version(unsigned int cmd, struct dm_ioctl __user *user)
return r;
}
-#define DM_PARAMS_KMALLOC 0x0001 /* Params alloced with kmalloc */
-#define DM_PARAMS_VMALLOC 0x0002 /* Params alloced with vmalloc */
#define DM_WIPE_BUFFER 0x0010 /* Wipe input buffer before returning from ioctl */
static void free_params(struct dm_ioctl *param, size_t param_size, int param_flags)
@@ -1679,10 +1677,7 @@ static void free_params(struct dm_ioctl *param, size_t param_size, int param_fla
if (param_flags & DM_WIPE_BUFFER)
memset(param, 0, param_size);
- if (param_flags & DM_PARAMS_KMALLOC)
- kfree(param);
- if (param_flags & DM_PARAMS_VMALLOC)
- vfree(param);
+ kvfree(param);
}
static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kernel,
@@ -1716,8 +1711,6 @@ static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kern
dmi = NULL;
if (param_kernel->data_size <= KMALLOC_MAX_SIZE) {
dmi = kmalloc(param_kernel->data_size, GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
- if (dmi)
- *param_flags |= DM_PARAMS_KMALLOC;
}
if (!dmi) {
@@ -1725,8 +1718,6 @@ static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kern
noio_flag = memalloc_noio_save();
dmi = __vmalloc(param_kernel->data_size, GFP_NOIO | __GFP_REPEAT | __GFP_HIGH | __GFP_HIGHMEM, PAGE_KERNEL);
memalloc_noio_restore(noio_flag);
- if (dmi)
- *param_flags |= DM_PARAMS_VMALLOC;
}
if (!dmi) {
--
1.9.1
On Mon, Apr 11 2016 at 11:14am -0400,
Sudip Mukherjee <[email protected]> wrote:
> We can use kvfree() instead of calling kfree() and vfree() based on
> if-else and param_flags. kvfree() will check the type of address and
> will call the respective function to free it.
> Additionally we can also remove the use of DM_PARAMS_KMALLOC and
> DM_PARAMS_VMALLOC.
>
> Signed-off-by: Sudip Mukherjee <[email protected]>
Have you actually tested htis? Last time I looked to do this it exposed
crashes. I don't have time to dig into this again right now but this is
_not_ as simple as this patch implies.
On Monday 11 April 2016 08:47 PM, Mike Snitzer wrote:
> On Mon, Apr 11 2016 at 11:14am -0400,
> Sudip Mukherjee <[email protected]> wrote:
>
>> We can use kvfree() instead of calling kfree() and vfree() based on
>> if-else and param_flags. kvfree() will check the type of address and
>> will call the respective function to free it.
>> Additionally we can also remove the use of DM_PARAMS_KMALLOC and
>> DM_PARAMS_VMALLOC.
>>
>> Signed-off-by: Sudip Mukherjee <[email protected]>
>
> Have you actually tested htis? Last time I looked to do this it exposed
> crashes. I don't have time to dig into this again right now but this is
> _not_ as simple as this patch implies.
>
No, it was just build tested. Is it possible to test it in qemu or kvm?
regards
sudip
On Mon, Apr 11, 2016 at 08:44:37PM +0530, Sudip Mukherjee wrote:
> We can use kvfree() instead of calling kfree() and vfree() based on
> if-else and param_flags. kvfree() will check the type of address and
> will call the respective function to free it.
> Additionally we can also remove the use of DM_PARAMS_KMALLOC and
> DM_PARAMS_VMALLOC.
>
> Signed-off-by: Sudip Mukherjee <[email protected]>
>
> ---
> drivers/md/dm-ioctl.c | 11 +----------
> 1 file changed, 1 insertion(+), 10 deletions(-)
>
> diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
> index 2adf81d..d5df3a5 100644
> --- a/drivers/md/dm-ioctl.c
> +++ b/drivers/md/dm-ioctl.c
> @@ -1670,8 +1670,6 @@ static int check_version(unsigned int cmd, struct dm_ioctl __user *user)
> return r;
> }
>
> -#define DM_PARAMS_KMALLOC 0x0001 /* Params alloced with kmalloc */
> -#define DM_PARAMS_VMALLOC 0x0002 /* Params alloced with vmalloc */
> #define DM_WIPE_BUFFER 0x0010 /* Wipe input buffer before returning from ioctl */
>
> static void free_params(struct dm_ioctl *param, size_t param_size, int param_flags)
> @@ -1679,10 +1677,7 @@ static void free_params(struct dm_ioctl *param, size_t param_size, int param_fla
> if (param_flags & DM_WIPE_BUFFER)
> memset(param, 0, param_size);
>
> - if (param_flags & DM_PARAMS_KMALLOC)
> - kfree(param);
> - if (param_flags & DM_PARAMS_VMALLOC)
> - vfree(param);
> + kvfree(param);
That won't work: param is not always allocated. See code path leading to
the "data_copied:" label in copy_params().
This is the second time this patch has been submitted. Maybe someone should
add a comment explaining why a conditional flag is needed.
Guenter
> }
>
> static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kernel,
> @@ -1716,8 +1711,6 @@ static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kern
> dmi = NULL;
> if (param_kernel->data_size <= KMALLOC_MAX_SIZE) {
> dmi = kmalloc(param_kernel->data_size, GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
> - if (dmi)
> - *param_flags |= DM_PARAMS_KMALLOC;
> }
>
> if (!dmi) {
> @@ -1725,8 +1718,6 @@ static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kern
> noio_flag = memalloc_noio_save();
> dmi = __vmalloc(param_kernel->data_size, GFP_NOIO | __GFP_REPEAT | __GFP_HIGH | __GFP_HIGHMEM, PAGE_KERNEL);
> memalloc_noio_restore(noio_flag);
> - if (dmi)
> - *param_flags |= DM_PARAMS_VMALLOC;
> }
>
> if (!dmi) {
On Mon, Apr 11 2016, Sudip Mukherjee <[email protected]> wrote:
> On Monday 11 April 2016 08:47 PM, Mike Snitzer wrote:
>> On Mon, Apr 11 2016 at 11:14am -0400,
>> Sudip Mukherjee <[email protected]> wrote:
>>
>>> We can use kvfree() instead of calling kfree() and vfree() based on
>>> if-else and param_flags. kvfree() will check the type of address and
>>> will call the respective function to free it.
>>> Additionally we can also remove the use of DM_PARAMS_KMALLOC and
>>> DM_PARAMS_VMALLOC.
>>>
>>> Signed-off-by: Sudip Mukherjee <[email protected]>
>>
>> Have you actually tested htis? Last time I looked to do this it exposed
>> crashes. I don't have time to dig into this again right now but this is
>> _not_ as simple as this patch implies.
>>
>
> No, it was just build tested. Is it possible to test it in qemu or kvm?
>
No need to test it, just read copy_params() and its caller,
ctl_ioctl(). The latter passes a stack buffer as param_kernel, and
copy_params() does
if (ioctl_flags & IOCTL_FLAGS_NO_PARAMS) {
dmi = param_kernel;
with dmi later returned via *param. So it is entirely possible that
free_params ends up calling neither kfree or vfree, since there's
nothing to free.
Rasmus