Re: [PATCH net] ice: Fix freeing uninitialized pointers

[Jesse Brandeburg]

On 3/21/2024 1:20 PM, Julia Lawall wrote:
> Does one prefer an initialization of null at the top of the function
> or an initialization to a meaningful value in the middle of the
> function ?

I think the latter.

There was a related patch explaining the direction, from Dan posted here:
https://lore.kernel.org/all/171097196970.1011049.9726486429680041876.stgit@xxxxxxxxxxxxxxxxxxxxxxxxx/

We had been having some internal discussions about use of __free(kfree) 
in the ice driver.

The gist of it is that we should instead be using inline declarations, 
which I also agree is a reasonable style for this. It more clearly shows 
the __free(kfree) and the allocation (kzalloc, kcalloc, etc) on the same 
(or virtually the same) line of code.

I'm curious if Jakub would dislike this less? Accept?

as an example:
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c 
b/drivers/net/ethernet/intel/ice/ice_common.c
index 88c86de82e09..822628d25b2f 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -1003,8 +1003,6 @@ static void ice_get_itr_intrl_gran(struct ice_hw *hw)
  */
 int ice_init_hw(struct ice_hw *hw)
 {
-       struct ice_aqc_get_phy_caps_data *pcaps __free(kfree) = NULL;
        void *mac_buf __free(kfree) = NULL;
        u16 mac_buf_len;
        int status;

@@ -1083,7 +1081,8 @@ int ice_init_hw(struct ice_hw *hw)
        if (status)
                goto err_unroll_sched;

-       pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
+       struct ice_aqc_get_phy_caps_data *pcaps __free(kfree) =
+               kzalloc(sizeof(*pcaps), GFP_KERNEL);
        if (!pcaps) {
                status = -ENOMEM;
                goto err_unroll_sched;

Any thoughts?