Re: [PATCH -mm 0/7] execns syscall and user namespace

["H. Peter Anvin" <hpa@zytor.com>]

Ulrich Drepper wrote:
> On 7/11/06, H. Peter Anvin <hpa@zytor.com> wrote:
>> > #define EXECVEF_NEWNS 0x00000100
>> > #define EXECVEF_NEWIPC        0x00000200
>> > #define EXECVEF_NEWUTS        0x00000400
>> > #define EXECVEF_NEWUSER       0x00000800
> 
> Yes on these.
> 
> 
>> If flags comes first, I would rather like to call it execfve(), or
>> perhaps execxve() ("extended") or execove() ("options").  execfve()
>> sounds like it executes a file descriptor (which would probably be
>> called fexecve()).
> 
> I think execfve is fine.
> 
> 
>> Perhaps more seriously, if we're adding more functionality already, it
>> should acquire -at functionality (execveat) and take a directory 
>> argument.
> 
> We have fexecve already.  Adding -at variants is probably not the best
> idea, it's confusing.  Note, that fexecve only takes a file
> descriptor, not a file descriptor plus file name.
> 
> The only reason I could see for changing this is thatfexecve depends
> on /proc.  But there is so much other functionality which won't work
> if /proc isn't mounted that I'd rank this low.  I'm fine with just
> adding execfve.

It seems to me to make a lot of sense to make it execveat(), then.  That 
way it would provide the equivalent functionality of both execve() and 
fexecve(), plus additional functionality.

	-hpa