Re: New monit web-interface

[Thomas Oppel]

Am Freitag, 12. Juli 2002 12:33 schrieb Martin Pala:
> Jan-Henrik Haukeland wrote:
> > Christian Hopp <address@hidden> writes:
> >>>I agree with Thomas, it is less secure when checksum will depend on
> >>>ctime. I think that solution outlined above (with configuration swith)
> >>>will be useful to allow sysadmin choose check for every cycle (more
> >>>security) or performance instead, such as:
> >>>
> >>>[set checksumAlways {true|false}]
> >>>
> >>>If not specified, true should be default (i think).
> >>>
> >>>
> >>>What do you think about it?
> >>
> >>A good point... I will take care of a "checksumAlways {true|false}"
> >>option!
> >
> > Maybe it could be an idea and check both ctime and mtime? I created
> > the following function before I planed to start the implementation:
> >
> >
> > new in monitor.h:
> >
> > #define MAX(x,y)           ((x) > (y) ? (x) : (y))
> >
> > in util.c:
> >
> > /**
> >  * Get a files last modified timestamp. This function returns the max
> >  * of either st_mtime or st_ctime. If the file does not exist or is
> >  * not a regular file FALSE is returned
> >  * @param file A file to stat
> >  * @return last modification time or FALSE if not found or not a regular
> > file */
> > long getchange_file(char *file) {
> >
> >   struct stat buf;
> >
> >   if(!stat(file, &buf))
> >       if(S_ISREG(buf.st_mode))
> >       return MAX(buf.st_mtime, buf.st_ctime);
> >
> >   return FALSE;
> >
> > }
>
> Problem is, that mtime can easily be modified (for example by unix
> 'touch' command). Its not big problem for intruder to mangle this value.
>
> Martin

Yep. time is alwyas a bad advisor. It will lead you in trouble sooner or 
later.
A hash on a bit order you can hardly fool.

Leppo