On Tue, Aug 15, 2017 at 09:49:47PM +0200, Michael Kerrisk (man-pages) wrote: > Hello Cyril, > > > On 08/06/2017 09:38 AM, Cyrill Gorcunov wrote: > > KCMP_EPOLL_TFD > > Check whether the file descriptor idx1 of process pid1 present in epoll(7) > > instance described by idx2 of process pid2. The argument idx2 is an > > address of structure where the target file is described. > > > > struct kcmp_epoll_slot { > > __u32 efd; > > __u32 tfd; > > __u64 toff; > > }; > > > > where efd is epoll file descriptor from epoll_create(2) call, tfd is a > > target file number, toff is a target file offset counted from zero. Sev‐ > > eral different targets may be registered with the same file descriptor > > number and setting specific offset helps to investigate each of them. > > Thanks. I've applied this patch. Is there some example program that > demonstrates the use of this feature? Well, not as its own. There are a series floating around for CRIU not merged yet since I need to rebase it on new CRIU release and resend, but there is a sample https://lists.openvz.org/pipermail/criu/2017-February/036041.html - the file added into epoll, then dup'ed and original is closed, so kcmp with dup'ed instance should match kcmp call. If you can wait I'll try to find some time on the week and rather update kernel selftests for this new kcmp mode and CC you on the patch. Sounds OK? -- To unsubscribe from this list: send the line "unsubscribe linux-api" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html