One of the more common cases of allocation size calculations is finding
the size of a structure that has a zero-sized array at the end, along
with memory for some number of elements for that array. For example:
struct foo {
int stuff;
void *entry[];
};
instance = kzalloc(sizeof(struct foo) + sizeof(void *) * count, GFP_KERNEL);
Instead of leaving these open-coded and prone to type mistakes, we can
now use the new struct_size() helper:
instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL);
This issue was detected with the help of Coccinelle.
Signed-off-by: Gustavo A. R. Silva <[email protected]>
---
drivers/net/wireless/intel/iwlwifi/mvm/d3.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/d3.c b/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
index 79bdae9..6960318 100644
--- a/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
+++ b/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
@@ -1769,8 +1769,7 @@ static void iwl_mvm_query_netdetect_reasons(struct iwl_mvm *mvm,
n_matches = 0;
}
- net_detect = kzalloc(sizeof(*net_detect) +
- (n_matches * sizeof(net_detect->matches[0])),
+ net_detect = kzalloc(struct_size(net_detect, matches, n_matches),
GFP_KERNEL);
if (!net_detect || !n_matches)
goto out_report_nd;
@@ -1785,8 +1784,7 @@ static void iwl_mvm_query_netdetect_reasons(struct iwl_mvm *mvm,
for (j = 0; j < SCAN_OFFLOAD_MATCHING_CHANNELS_LEN; j++)
n_channels += hweight8(fw_match->matching_channels[j]);
- match = kzalloc(sizeof(*match) +
- (n_channels * sizeof(*match->channels)),
+ match = kzalloc(struct_size(match, channels, n_channels),
GFP_KERNEL);
if (!match)
goto out_report_nd;
--
2.7.4
On Thu, Aug 23, 2018 at 6:15 PM, Gustavo A. R. Silva
<[email protected]> wrote:
> One of the more common cases of allocation size calculations is finding
> the size of a structure that has a zero-sized array at the end, along
> with memory for some number of elements for that array. For example:
>
> struct foo {
> int stuff;
> void *entry[];
> };
>
> instance = kzalloc(sizeof(struct foo) + sizeof(void *) * count, GFP_KERNEL);
>
> Instead of leaving these open-coded and prone to type mistakes, we can
> now use the new struct_size() helper:
>
> instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL);
>
> This issue was detected with the help of Coccinelle.
>
> Signed-off-by: Gustavo A. R. Silva <[email protected]>
Reviewed-by: Kees Cook <[email protected]>
-Kees
> ---
> drivers/net/wireless/intel/iwlwifi/mvm/d3.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/d3.c b/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
> index 79bdae9..6960318 100644
> --- a/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
> +++ b/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
> @@ -1769,8 +1769,7 @@ static void iwl_mvm_query_netdetect_reasons(struct iwl_mvm *mvm,
> n_matches = 0;
> }
>
> - net_detect = kzalloc(sizeof(*net_detect) +
> - (n_matches * sizeof(net_detect->matches[0])),
> + net_detect = kzalloc(struct_size(net_detect, matches, n_matches),
> GFP_KERNEL);
> if (!net_detect || !n_matches)
> goto out_report_nd;
> @@ -1785,8 +1784,7 @@ static void iwl_mvm_query_netdetect_reasons(struct iwl_mvm *mvm,
> for (j = 0; j < SCAN_OFFLOAD_MATCHING_CHANNELS_LEN; j++)
> n_channels += hweight8(fw_match->matching_channels[j]);
>
> - match = kzalloc(sizeof(*match) +
> - (n_channels * sizeof(*match->channels)),
> + match = kzalloc(struct_size(match, channels, n_channels),
> GFP_KERNEL);
> if (!match)
> goto out_report_nd;
> --
> 2.7.4
>
--
Kees Cook
Pixel Security
On Thu, 2018-08-23 at 20:03 -0700, Kees Cook wrote:
> On Thu, Aug 23, 2018 at 6:15 PM, Gustavo A. R. Silva
> <[email protected]> wrote:
> > One of the more common cases of allocation size calculations is finding
> > the size of a structure that has a zero-sized array at the end, along
> > with memory for some number of elements for that array. For example:
> >
> > struct foo {
> > int stuff;
> > void *entry[];
> > };
Question for Gustavo.
Did you find any existing instances that are miscalculated?
I believe there are some cases like:
size = sizeof(struct foo) + count * sizeof(something);
ptr = kmalloc(size);
memset(ptr + sizeof(struct foo), 0, size - sizeof(struct foo));
where something could go wrong and not be detected.
Hi Joe,
On 8/23/18 11:33 PM, Joe Perches wrote:
> On Thu, 2018-08-23 at 20:03 -0700, Kees Cook wrote:
>> On Thu, Aug 23, 2018 at 6:15 PM, Gustavo A. R. Silva
>> <[email protected]> wrote:
>>> One of the more common cases of allocation size calculations is finding
>>> the size of a structure that has a zero-sized array at the end, along
>>> with memory for some number of elements for that array. For example:
>>>
>>> struct foo {
>>> int stuff;
>>> void *entry[];
>>> };
>
> Question for Gustavo.
>
> Did you find any existing instances that are miscalculated?
>
I found the following bug:
https://lore.kernel.org/patchwork/patch/977357/
> I believe there are some cases like:
>
> size = sizeof(struct foo) + count * sizeof(something);
> ptr = kmalloc(size);
> memset(ptr + sizeof(struct foo), 0, size - sizeof(struct foo));
>
> where something could go wrong and not be detected.
>
It might be worth it to write a Coccinelle script for this.
--
Gustavo
On Thu, 2018-08-23 at 20:15 -0500, Gustavo A. R. Silva wrote:
> One of the more common cases of allocation size calculations is finding
> the size of a structure that has a zero-sized array at the end, along
> with memory for some number of elements for that array. For example:
>
> struct foo {
> int stuff;
> void *entry[];
> };
>
> instance = kzalloc(sizeof(struct foo) + sizeof(void *) * count, GFP_KERNEL);
>
> Instead of leaving these open-coded and prone to type mistakes, we can
> now use the new struct_size() helper:
>
> instance = kzalloc(struct_size(instance, entry, count), GFP_KERNEL);
>
> This issue was detected with the help of Coccinelle.
>
> Signed-off-by: Gustavo A. R. Silva <[email protected]>
> ---
> drivers/net/wireless/intel/iwlwifi/mvm/d3.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/d3.c b/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
> index 79bdae9..6960318 100644
> --- a/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
> +++ b/drivers/net/wireless/intel/iwlwifi/mvm/d3.c
> @@ -1769,8 +1769,7 @@ static void iwl_mvm_query_netdetect_reasons(struct iwl_mvm *mvm,
> n_matches = 0;
> }
>
> - net_detect = kzalloc(sizeof(*net_detect) +
> - (n_matches * sizeof(net_detect->matches[0])),
> + net_detect = kzalloc(struct_size(net_detect, matches, n_matches),
> GFP_KERNEL);
> if (!net_detect || !n_matches)
> goto out_report_nd;
> @@ -1785,8 +1784,7 @@ static void iwl_mvm_query_netdetect_reasons(struct iwl_mvm *mvm,
> for (j = 0; j < SCAN_OFFLOAD_MATCHING_CHANNELS_LEN; j++)
> n_channels += hweight8(fw_match->matching_channels[j]);
>
> - match = kzalloc(sizeof(*match) +
> - (n_channels * sizeof(*match->channels)),
> + match = kzalloc(struct_size(match, channels, n_channels),
> GFP_KERNEL);
> if (!match)
> goto out_report_nd;
Thanks! I applied this to our internal tree and it will reach the
mainline following our normal process.
--
Cheers,
Luca.