On Monday 12 June 2023 23:52:30 Andy Shevchenko wrote: > On Mon, Jun 12, 2023 at 07:52:05PM +0200, Pali Rohár wrote: > > On Monday 12 June 2023 12:02:50 Michal Wilczynski wrote: > > > Currently rbtn_add() in case of failure is leaking resources. Fix this > > > by adding a proper rollback. While at it, remove unnecessary assignment > > > of NULL to device->driver_data and unnecessary whitespace, plus add a > > > break for the default case in a switch. > > ... > > > Hello! I'm looking at rbtn_add() function and there is also code: > > > > rbtn_data = devm_kzalloc(&device->dev, sizeof(*rbtn_data), GFP_KERNEL); > > if (!rbtn_data) > > return -ENOMEM; > > > > which is called after rbtn_acquire(). So it looks like when kzalloc > > fails then there is another leak... > > Side note: In that case we would need a devm wrapper on acquire call. Does it makes sense to invest time and more resources for these fixes? Driver is not used on new Dell machines, so I would not expect new users (instead less users, if they start upgrading HW to new Dell machines). Simple fix for this issue: Just move devm_kzalloc() call before rbtn_acquire(true). And call cleanup rbtn_acquire(false) exactly like Michal did in this patch. I think that this should be enough, should cover all failure paths and does not require to introduce new code or new design, which should be properly tested for no regression. What do you think?