Re: [PATCH BlueZ] core: Fix walking the list while removing elements

[Johan Hedberg <johan.hedberg@gmail.com>]

Hi Lucas,

On Thu, Oct 04, 2012, Lucas De Marchi wrote:
> If we are walking a GSList and remove the element we are pointing to,
> the next iteration g_slist_next() will access previously freed
> memory.
> ---
> 
> This was caught only by inspecting the code. I don't know why valgrind
> didn't complain about accessing previously freed memory region.
> 
>  src/device.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/src/device.c b/src/device.c
> index c659164..6150963 100644
> --- a/src/device.c
> +++ b/src/device.c
> @@ -1498,7 +1498,7 @@ static void device_remove_profiles(struct btd_device *device, GSList *uuids)
>  	if (records)
>  		sdp_list_free(records, (sdp_free_func_t) sdp_record_free);
>  
> -	for (l = device->profiles; l != NULL; l = g_slist_next(l)) {
> +	for (l = device->profiles; l != NULL;) {
>  		struct btd_profile *profile = l->data;
>  		GSList *probe_uuids;
>  
> @@ -1506,9 +1506,11 @@ static void device_remove_profiles(struct btd_device *device, GSList *uuids)
>  								device->uuids);
>  		if (probe_uuids != NULL) {
>  			g_slist_free(probe_uuids);
> +			l = l->next;
>  			continue;
>  		}
>  
> +		l = l->next;
>  		profile->device_remove(profile, device);
>  		device->profiles = g_slist_remove(device->profiles, profile);
>  	}

Thanks for catching this, however could you fix it the same way most
other similar loops in the code-base do it, i.e. add a GSList *next
helper variable:

	GSList *l, *next;

	for (l = device->profiles; l != NULL; l = next) {
		<variable declarations>

		next = l->next;

		<rest of the loop code>
	}

Johan