nss_files testing

[Florian Weimer]

I'm attaching a preview patch for testing nss_files.  On recent kernels, 
with user namespaces enabled, the test can run without special 
privileges.  There is also direct chroot support for systems which lack 
user namespaces.

One tricky bit is how to get the device nodes in /dev.  I use a 
recursive bind mount for this (non-recursive outside of a namespace, due 
to bad devpts interactions).  This means that the code is currently 
Linux-specific.  I could probably add something that recreates critical 
device nodes (null, zero, random, urandom) when running as root, then 
the requirement for the bind mount would go away.

Some things need cleanup.  I want to submit this for real once I have 
some experience with adding nss_db support (which will reuse the chroot 
setup code from tst-nss-files-aux, moving it to tst-nss-files).  Longer 
term, this approach should also work for testing nscd.

tst-nss-files-enumeration currently fails (bug 18991).

Florian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: nss-files.patch
Type: text/x-patch
Size: 53819 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20160415/bb00a471/attachment.bin>