[LTP] [PATCH 2/2] needs_root: Add setting of needs_root

[zhaogongyi]

Hi,

> 
> Hi all,
> 
> > Hi,
> 
> 
> > > > > > Hi Petr,
> 
> > > > > > It is failed on my system:
> 
> > > > > > sh-4.4$ ./msgget03
> > > > > > tst_test.c:1535: TINFO: Timeout per run is 0h 00m 30s
> > > > > > msgget03.c:42: TINFO: Current environment 0 message queues
> are
> > > > > already
> > > > > > in use
> > > > > > msgget03.c:45: TBROK: Failed to open FILE
> > > '/proc/sys/kernel/msgmni'
> > > > > > for writing: EACCES (13)
> 
> > > > > Ah, b740bfac5 ("msgget03: Set custom queue limit") causes the
> > > > > need for root. IMHO this one is valid.
> 
> > > > > I'd suggest to remove needs_root for needs_device and
> > > mount_device
> > > > > now (i.e. send v2).
> 
> > > > Just remove the patch 1 and keep adding needs_root in patch 2?
> 
> > > If you haven't added it to any test which would have needs_device or
> > > mount_device, then v2 can be kept.
> 
> > > But process_vm_readv02 works for me (on master, on Debian) without
> > > root:
> 
> > > $ ./process_vm_readv02
> > > tst_test.c:1526: TINFO: Timeout per run is 0h 00m 30s
> > > process_vm_readv02.c:32: TINFO: child 0: memory allocated and
> > > initialized
> > > process_vm_readv02.c:49: TINFO: child 1: reading string from same
> > > memory location
> > > process_vm_readv02.c:61: TPASS: expected string received
> 
> > > Summary:
> > > passed   1
> > > failed   0
> > > broken   0
> > > skipped  0
> > > warnings 0
> 
> > > Also process_vm_readv03, process_vm_writev02 (I stopped checking
> > > now).
> 
> > > Why you need it?
> 
> 
> 
> > On my system, these testcases failed and report EPERM:
> 
> > sh-4.4$ ./process_vm_readv02
> OK, your system is Alpine (it can shorten our discussion next time if you
> report you have problems on musl - i.e. which kernel and libc + versions
> you use).
> 
> Well, we should first check if this is not a bug in musl or Alpine before
> requiring root.
> 
> It'd be interesting to know whether
> tools/testing/selftests/vm/memfd_secret.c
> and tools/testing/selftests/x86/test_vsyscall.c also require root.
> 
> > tst_test.c:1528: TINFO: Timeout per run is 0h 00m 30s
> > process_vm_readv02.c:32: TINFO: child 0: memory allocated and
> > initialized
> > process_vm_readv02.c:49: TINFO: child 1: reading string from same
> > memory location
> > process_vm_readv02.c:55: TBROK: process_vm_readv: EPERM
> > process_vm_readv02.c:107: TFAIL: child 1: exited with 2
> 

Because of the tests bind06/msgget03/sendto03/setsockopt0[5-9]/swapon01 don't reference to
mount device or different outcome on kinds of kernel versions, should I resubmit a new patch to fix them first?

In fact, I have no a system with kernel version 4.15 that can trace what has happened for tests process_vm_* now.

Regards,
Gongyi