Re: [PATCH net-next 4/6] ice: remove null checks before devm_kfree() calls

[Maciej Fijalkowski <maciej.fijalkowski@intel.com>]

On Fri, Jun 23, 2023 at 03:21:30PM +0200, Przemek Kitszel wrote:
> On 6/23/23 13:13, Maciej Fijalkowski wrote:
> > On Fri, Jun 23, 2023 at 12:44:29PM +0200, Przemek Kitszel wrote:
> > > On 6/23/23 12:23, Maciej Fijalkowski wrote:
> > > > On Thu, Jun 22, 2023 at 11:35:59AM -0700, Tony Nguyen wrote:
> > > > > From: Przemek Kitszel <przemyslaw.kitszel@intel.com>
> > > > > 
> > > > > We all know they are redundant.
> > > > 
> > > > Przemek,
> > > > 
> > > > Ok, they are redundant, but could you also audit the driver if these devm_
> > > > allocations could become a plain kzalloc/kfree calls?
> > > 
> > > Olek was also motivating such audit :)
> > > 
> > > I have some cases collected with intention to send in bulk for next window,
> > > list is not exhaustive though.
> > 
> > When rev-by count tag would be considered too much? I have a mixed
> > feelings about providing yet another one, however...
> > 
> > > 
> > > > 
> > > > > 
> > > > > Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> > > > > Reviewed-by: Michal Wilczynski <michal.wilczynski@intel.com>
> > > > > Reviewed-by: Simon Horman <simon.horman@corigine.com>
> > > > > Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
> > > > > Tested-by: Arpana Arland <arpanax.arland@intel.com> (A Contingent worker at Intel)
> > > > > Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > > ---
> > > > >    drivers/net/ethernet/intel/ice/ice_common.c   |  6 +--
> > > > >    drivers/net/ethernet/intel/ice/ice_controlq.c |  3 +-
> > > > >    drivers/net/ethernet/intel/ice/ice_flow.c     | 23 ++--------
> > > > >    drivers/net/ethernet/intel/ice/ice_lib.c      | 42 +++++++------------
> > > > >    drivers/net/ethernet/intel/ice/ice_sched.c    | 11 ++---
> > > > >    drivers/net/ethernet/intel/ice/ice_switch.c   | 19 +++------
> > > > >    6 files changed, 29 insertions(+), 75 deletions(-)
> > > > > 
> > > > > diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
> > > > > index eb2dc0983776..6acb40f3c202 100644
> > > > > --- a/drivers/net/ethernet/intel/ice/ice_common.c
> > > > > +++ b/drivers/net/ethernet/intel/ice/ice_common.c
> > > > > @@ -814,8 +814,7 @@ static void ice_cleanup_fltr_mgmt_struct(struct ice_hw *hw)
> > > > >    				devm_kfree(ice_hw_to_dev(hw), lst_itr);
> > > > >    			}
> > > > >    		}
> > > > > -		if (recps[i].root_buf)
> > > > > -			devm_kfree(ice_hw_to_dev(hw), recps[i].root_buf);
> > > > > +		devm_kfree(ice_hw_to_dev(hw), recps[i].root_buf);
> > > > >    	}
> > > > >    	ice_rm_all_sw_replay_rule_info(hw);
> > > > >    	devm_kfree(ice_hw_to_dev(hw), sw->recp_list);
> > > > > @@ -1011,8 +1010,7 @@ static int ice_cfg_fw_log(struct ice_hw *hw, bool enable)
> > > > >    	}
> > > > >    out:
> > > > > -	if (data)
> > > > > -		devm_kfree(ice_hw_to_dev(hw), data);
> > > > > +	devm_kfree(ice_hw_to_dev(hw), data);
> > > > >    	return status;
> > > > >    }
> > > > > diff --git a/drivers/net/ethernet/intel/ice/ice_controlq.c b/drivers/net/ethernet/intel/ice/ice_controlq.c
> > > > > index 385fd88831db..e7d2474c431c 100644
> > > > > --- a/drivers/net/ethernet/intel/ice/ice_controlq.c
> > > > > +++ b/drivers/net/ethernet/intel/ice/ice_controlq.c
> > > > > @@ -339,8 +339,7 @@ do {									\
> > > > >    		}							\
> > > > >    	}								\
> > > > >    	/* free the buffer info list */					\
> > > > > -	if ((qi)->ring.cmd_buf)						\
> > > > > -		devm_kfree(ice_hw_to_dev(hw), (qi)->ring.cmd_buf);	\
> > > > > +	devm_kfree(ice_hw_to_dev(hw), (qi)->ring.cmd_buf);		\
> > > > >    	/* free DMA head */						\
> > > > >    	devm_kfree(ice_hw_to_dev(hw), (qi)->ring.dma_head);		\
> > > > >    } while (0)
> > > > > diff --git a/drivers/net/ethernet/intel/ice/ice_flow.c b/drivers/net/ethernet/intel/ice/ice_flow.c
> > > > > index ef103e47a8dc..85cca572c22a 100644
> > > > > --- a/drivers/net/ethernet/intel/ice/ice_flow.c
> > > > > +++ b/drivers/net/ethernet/intel/ice/ice_flow.c
> > > > > @@ -1303,23 +1303,6 @@ ice_flow_find_prof_id(struct ice_hw *hw, enum ice_block blk, u64 prof_id)
> > > > >    	return NULL;
> > > > >    }
> > > > > -/**
> > > > > - * ice_dealloc_flow_entry - Deallocate flow entry memory
> > > > > - * @hw: pointer to the HW struct
> > > > > - * @entry: flow entry to be removed
> > > > > - */
> > > > > -static void
> > > > > -ice_dealloc_flow_entry(struct ice_hw *hw, struct ice_flow_entry *entry)
> > > > > -{
> > > > > -	if (!entry)
> > > > > -		return;
> > > > > -
> > > > > -	if (entry->entry)
> > 
> > ...would you be able to point me to the chunk of code that actually sets
> > ice_flow_entry::entry? because from a quick glance I can't seem to find
> > it.
> 
> Simon was asking very similar question [1],
> albeit "where is the *check* for entry not being null?" (not set),
> and it is just above the default three lines of context provided by git
> (pasted below for your convenience, [3])
> 
> To answer, "where it's set?", see ice_flow_add_entry(), [2].

I was referring to 'entry' member from ice_flow_entry struct. You're
pointing me to init of whole ice_flow_entry.

I am trying to say that if ice_flow_entry::entry is never set, then
probably it could be removed from struct.

> 
> [1] https://lore.kernel.org/netdev/ZHb5AIgL5SzBa5FA@corigine.com/
> [2] https://elixir.bootlin.com/linux/v6.4-rc7/source/drivers/net/ethernet/intel/ice/ice_flow.c#L1632
> 
> --
> 
> BTW, is there any option to add some of patch generation options (like,
> context size, anchored lines, etc), that there are my disposal locally, but
> in a way, that it would not be lost after patch is applied to one tree
> (Tony's) and then send again (here)?
> (My assumption is that Tony is (re)generating patches from git (opposed to
> copy-pasting+decorating of what he has received from me).
> 
> 
> 
> > 
> > > > > -		devm_kfree(ice_hw_to_dev(hw), entry->entry);
> > > > > -
> > > > > -	devm_kfree(ice_hw_to_dev(hw), entry);
> > > > > -}
> > > > > -
> > > > >    /**
> > > > >     * ice_flow_rem_entry_sync - Remove a flow entry
> > > > >     * @hw: pointer to the HW struct
> > > > > @@ -1335,7 +1318,8 @@ ice_flow_rem_entry_sync(struct ice_hw *hw, enum ice_block __always_unused blk,
> 
> [3] More context would include following:
> 
>          if (!entry)
>                  return -EINVAL;
> 
> > > > >    	list_del(&entry->l_entry);
> > > > > -	ice_dealloc_flow_entry(hw, entry);
> > > > > +	devm_kfree(ice_hw_to_dev(hw), entry->entry);
> > > > > +	devm_kfree(ice_hw_to_dev(hw), entry);
> > > > >    	return 0;
> > > > >    }
> > > > > @@ -1662,8 +1646,7 @@ ice_flow_add_entry(struct ice_hw *hw, enum ice_block blk, u64 prof_id,
> > > > >    out:
> > > > >    	if (status && e) {
> > > > > -		if (e->entry)
> > > > > -			devm_kfree(ice_hw_to_dev(hw), e->entry);
> > > > > +		devm_kfree(ice_hw_to_dev(hw), e->entry);
> > > > >    		devm_kfree(ice_hw_to_dev(hw), e);
> > > > >    	}
> > 
> > (...)
>