13 октября 2014 г. 13:53:48 KRAT, Rob Landley <[email protected]> пишет: >On 10/13/14 00:29, Рысь wrote: >> 13 октября 2014 г. 8:51:13 KRAT, Rob Landley <[email protected]> пишет: >>> My question is: should I _also_ include the old raw socket support >(as >>> a >>> fallback? With a config option?) or is it good enough to just have >this >>> and not support ping on 2.6 kernels? (It's easy enough to make it >>> compile even when it does work, or to have a compile-time probe to >>> switch it off in defconfig...) >>> >> >> Will ping work when invoked by root or not in that case? > >The new way does not require ping to run as root. > >> The old and always worked way is prefferable IMO. Is there a reason >to change? >> I, for example, use ping (and other such tools) with a universal >setuid wrapper >> which (re)execs it as "user,root:user,user". I do not want any >breakage here in future. > >Toybox has its own suid handling, which requires the suid bit on the >toybox executable. (Or you can build a toybox executable with just the >suid commands and a second one with the non suid commands, and point >the >symlinks appropriately.) > >The thing is, Linux added an API that lets you do the same thing >without >requiring root access to do so. Seems like an improvement, especially >since toybox isn't always installed with the suid bit set. (So you have >to run things _as_ root to use the stuff requiring root.)
So toybox ping will work regardless of kernel version, even 2.6.1x? > >Rob >_______________________________________________ >Toybox mailing list >[email protected] >http://lists.landley.net/listinfo.cgi/toybox-landley.net _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
