On Tue, 2019-07-23 at 15:52 -0500, Julia Lawall wrote: > On Mon, 22 Jul 2019, Joe Perches wrote: > > I just sent a patch to add yet another string copy mechanism. > > > > This could help avoid misuses of strscpy and strlcpy like this > > patch set: > > > > https://lore.kernel.org/lkml/cover.1562283944.git....@perches.com/T/ > > > > A possible cocci script to do conversions could be: > > > > $ cat str.cpy.cocci > > @@ > > expression e1; > > expression e2; > > @@ > > > > - strscpy(e1, e2, sizeof(e1)) > > + stracpy(e1, e2) > > > > @@ > > expression e1; > > expression e2; > > @@ > > > > - strlcpy(e1, e2, sizeof(e1)) > > + stracpy(e1, e2) > > > > This obviously does not match the style of all the > > scripts/coccinelle cocci files, but this might be > > something that could be added improved and added. > > > > This script produces: > > > > $ spatch --in-place -sp-file str.cpy.cocci . > > $ git checkout tools/ > > $ git diff --shortstat > > 958 files changed, 2179 insertions(+), 2655 deletions(-) > > > > The remainder of strlcpy and strscpy uses in the > > kernel would generally have a form like: > > > > strlcpy(to, from, DEFINE) > > > > where DEFINE is the specified size of to > > > > Could the cocci script above be updated to find > > and correct those styles as well? > > I guess it would depend on what "to" is and what DEFINE expands into. For > example, in cpuidle-powernv.c, I see: > > strlcpy(powernv_states[index].name, name, CPUIDLE_NAME_LEN); > > and by poking around I see: > > struct cpuidle_state { > char name[CPUIDLE_NAME_LEN]; > char desc[CPUIDLE_DESC_LEN]; > ... > };
Yes, ideally this case would not modify the #define for the length but adapt the strlcpy(,,DEFINE) There are a lot of these in drivers/hwmon using I2C_NAME_SIZE. > I will look into it. Thanks.
