Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> writes: > On Fri, May 18, 2012 at 9:08 AM, Linus Torvalds > <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: >> >> I would suggest just moving the i_mode initialization from >> proc_fd_instantiate() into the revalidate function that we already >> have, and that already fixes up i_uid/i_gid etc. Attached is a TOTALLY >> UNTESTED patch that does this, and actually seems to simplify things >> in the process. > > Ok, so it's now "tested" in the sense that it works for me, and fixes > both my and your test-cases. > > Which doesn't mean that it is bug-free of course, but it does seem to > be a sane patch that actually cleans things up. > > I'll delay committing it in case somebody hollers, but I think I'll > mark it for stable too since this issue seems age-old, and the fix > looks good. > > Ack/nak? Can anybody find anything wrong in that patch? Acked-by: "Eric W. Biederman" <ebiederm@xxxxxxxxxxxx> It looks reasonable. I am a tad leery of the d_add without setting inode->i_mode but the dcache lookup will have to call revalidate before we access the inode so I don't expect a problem there. Eric -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
