On Mon, Sep 07, 2020 at 12:28:05PM +0200, Bartosz Golaszewski wrote: > On Mon, Sep 7, 2020 at 12:19 PM Andy Shevchenko > <andy.shevchenko@xxxxxxxxx> wrote: > > > > On Mon, Sep 7, 2020 at 1:05 PM Bartosz Golaszewski > > <bgolaszewski@xxxxxxxxxxxx> wrote: > > > On Fri, Sep 4, 2020 at 6:35 PM Andy Shevchenko > > > <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote: > > > > On Fri, Sep 04, 2020 at 05:45:27PM +0200, Bartosz Golaszewski wrote: > > > > > From: Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx> > > > > > > Doesn't mm/util.c provides us something like this? > > > > strndup_user()? > > > > > > > > > > Yes, there's both strndup_user() as well as strncpy_from_user(). The > > > problem is that they rely on the strings being NULL-terminated. This > > > is not guaranteed for debugfs file_operations write callbacks. We need > > > some helper that takes the minimum of bytes provided by userspace and > > > the buffer size and figure out how many bytes to actually copy IMO. > > > > Wouldn't this [1] approach work? > > > > [1]: https://elixir.bootlin.com/linux/v5.9-rc3/source/arch/x86/kernel/cpu/mtrr/if.c#L93 > > > > Sure, but this is pretty much what I do in getline_from_user(). If > anything we should port mtrr_write() to using getline_from_user() once > it's available upstream, no? But you may provide getline_from_user() as inline in the same header where strncpy_from_user() is declared. It will be like 3 LOCs? -- With Best Regards, Andy Shevchenko
