On 08/15, Zach L wrote: > > On 08/14/2013 10:50 AM, Oleg Nesterov wrote: > > On 08/14, Zach Levis wrote: > >> > > Honestly, I dislike this version even more, sorry. The patch becomes > > much more complex, and and it is still not clear to me why do we want > > these complications. > > > It's a larger patch but the majority of the increase is from is > splitting the binfmt initialization code into a separate function to > address the issue you brought up where the state of the binprm was not > entirely restored I understand the reason. But I do not understand the value. IMHO, the problem this patch tries to fix falls into the "don't do this" category and doesn't worth the trouble. > [snip] This certainly answers my question you snipped ;) > > And btw, if we want this, then why we only do this if recursion_depth == 0? > > Just condider '#!/path-to-the-binary-which-wants-this-patch". > Unless recursion_depth is 0, there could be a binfmt in between that > would expect its changes to the binprm to remain in effect in lower > handlers, so even with your example My point was, this doesn't fix the same problem if depth != 0. But yes, "depth > 0" can't simply do init_bprm(). > > And again, the patch (afaics) translates -ELOOP into -ENOEXEC on failure, > > not good. > it doesn't do that, It does, afaics. Just suppose that -ELOOP comes from load_script(). We restore everything and call the next handler which returns ENOEXEC. And at first glance v5 does the same. Oleg. -- 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
