Re: [linux-next:master 4744/6155] net/dcb/dcbnl.c:1230:1: warning: the frame size of 1244 bytes is larger than 1024 bytes

[Petr Machata <petrm@nvidia.com>]

<Daniel.Machon@microchip.com> writes:

>> kernel test robot <lkp@intel.com> writes:
>> 
>> > tree:   https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
>> > head:   d8e87774068af213ab5b058b1b114dc397b577aa
>> > commit: 6182d5875c330a5a611687caa05f47752455720c [4744/6155] net: dcb: add new apptrust attribute
>> > config: i386-buildonly-randconfig-r006-20221107
>> > compiler: gcc-11 (Debian 11.3.0-8) 11.3.0
>> > reproduce (this is a W=1 build):
>> >         # https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?id=6182d5875c330a5a611687caa05f47752455720c
>> >         git remote add linux-next https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
>> >         git fetch --no-tags linux-next master
>> >         git checkout 6182d5875c330a5a611687caa05f47752455720c
>> >         # save the config file
>> >         mkdir build_dir && cp config build_dir/.config
>> >         make W=1 O=build_dir ARCH=i386 SHELL=/bin/bash net/dcb/
>> >
>> > If you fix the issue, kindly add following tag where applicable
>> > | Reported-by: kernel test robot <lkp@intel.com>
>> >
>> > All warnings (new ones prefixed by >>):
>> >
>> >    net/dcb/dcbnl.c: In function 'dcbnl_ieee_fill':
>> >>> net/dcb/dcbnl.c:1230:1: warning: the frame size of 1244 bytes is larger than 1024 bytes [-Wframe-larger-than=]
>> 
>> The on-stack array that was added is pushing things:
>> 
>> > 6182d5875c330a Daniel Machon   2022-11-01  1170       if (ops->dcbnl_getapptrust) {
>> > 6182d5875c330a Daniel Machon   2022-11-01  1171               u8 selectors[IEEE_8021QAZ_APP_SEL_MAX + 1] = {0};
>> > 6182d5875c330a Daniel Machon   2022-11-01  1172               int nselectors, i;
>> > 6182d5875c330a Daniel Machon   2022-11-01  1173
>> > 6182d5875c330a Daniel Machon   2022-11-01  1174               apptrust = nla_nest_start(skb, DCB_ATTR_DCB_APP_TRUST_TABLE);
>> > 6182d5875c330a Daniel Machon   2022-11-01  1175               if (!apptrust)
>> > 6182d5875c330a Daniel Machon   2022-11-01  1176                       return -EMSGSIZE;
>> > 6182d5875c330a Daniel Machon   2022-11-01  1177
>> > 6182d5875c330a Daniel Machon   2022-11-01  1178               err = ops->dcbnl_getapptrust(netdev, selectors, &nselectors);
>> > 6182d5875c330a Daniel Machon   2022-11-01  1179               if (!err) {
>> > 6182d5875c330a Daniel Machon   2022-11-01  1180                       for (i = 0; i < nselectors; i++) {
>> > 6182d5875c330a Daniel Machon   2022-11-01  1181                               enum ieee_attrs_app type =
>> > 6182d5875c330a Daniel Machon   2022-11-01  1182                                       dcbnl_app_attr_type_get(selectors[i]);
>> > 6182d5875c330a Daniel Machon   2022-11-01  1183                               err = nla_put_u8(skb, type, selectors[i]);
>> > 6182d5875c330a Daniel Machon   2022-11-01  1184                               if (err) {
>> > 6182d5875c330a Daniel Machon   2022-11-01  1185                                       nla_nest_cancel(skb, apptrust);
>> > 6182d5875c330a Daniel Machon   2022-11-01  1186                                       return err;
>> > 6182d5875c330a Daniel Machon   2022-11-01  1187                               }
>> > 6182d5875c330a Daniel Machon   2022-11-01  1188                       }
>> > 6182d5875c330a Daniel Machon   2022-11-01  1189               }
>> > 6182d5875c330a Daniel Machon   2022-11-01  1190
>> > 6182d5875c330a Daniel Machon   2022-11-01  1191               nla_nest_end(skb, apptrust);
>> > 6182d5875c330a Daniel Machon   2022-11-01  1192       }
>> 
>> I think this should either be converted to dynamic allocation, or the
>> array size should be reduced to number of unique selectors, instead of
>> the value of the highest selector.
>
> Yep. Dynamic allocation sounds good to me. Alternatively, could we move
> the getapptrust part to a separate noinline function?

I think the reasoning behind the warning is that functions with large
frames consume too much stack space, which is scarce in kernel. Moving
the logic to a separate function might silence the warning, but wouldn't
make the issue go away.

Extracting the logic to a separate function might still make sense,
because it's going to make it easier to manage the dynamic allocation,
but is not enough on its own.