On Wed, 28 Nov 2018 at 05:05, Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote: > > On Mon, Nov 26, 2018 at 4:45 AM Lorenz Bauer <lmb@xxxxxxxxxxxxxx> wrote: > > > > That's what I had initially, but that makes re-using test_attr really > > awkward. Either > > you need to reset data_out_size before every call because it is used > > to return the > > buffer size, > > I think that is exactly what the user of the interface would want to do. > Why would anyone keep reusing the same test_attr on multiple calls > into the kernel without changing the fields? Basically, you can only change the input part without having to reset data_size_out to sizeof(buffer). Not a big deal, I'll change it. > > > It also means > > we can't take a const struct attr, which is contrary to the other > > xattr functions which > > already exist. > > I don't see an issue with that. > > > I think actually inspecting the required size of the output buffer > > will be a rare > > occurrence, so making the user jump through the hoop of a pointer doesn't seem > > too onerous. > > I think the opposite is the case. > If the output buffer is provided the test will be comparing it > to expected value. Yeah, I wasn't thinking too hard on this one, sorry. User space needs to check where the end of the buffer is. -- Lorenz Bauer | Systems Engineer 25 Lavington St., London SE1 0NZ www.cloudflare.com
