Re: [PATCH] update ctime and mtime for mmaped write

[Miklos Szeredi]

> >>> Inspired by Peter Staubach's patch and the resulting comments.
> >>>
> >>>   
> >>>       
> >> An updated version of the original patch was submitted to LKML
> >> yesterday...  :-)
> >>     
> >
> > Strange coincidence :)
> >
> >   
> >>>  		file = vma->vm_file;
> >>>  		start = vma->vm_end;
> >>> +		mapping_update_time(file);
> >>>  		if ((flags & MS_SYNC) && file &&
> >>>  				(vma->vm_flags & VM_SHARED)) {
> >>>  			get_file(file);
> >>>   
> >>>       
> >> It seems to me that this might lead to file times being updated for
> >> non-MAP_SHARED mappings.
> >>     
> >
> > In theory no, because the COW-ed pages become anonymous and are not
> > part of the original mapping any more.
> >
> >   
> 
> I must profess to having a incomplete understanding of all of this
> support, but then why would it be necessary to test VM_SHARED at
> this point in msync()?

That's basically just an optimization.  If it wasn't there, then data
from a another (shared) mapping could be written back, which is not
wrong, but not required either.

> I ran into problems early on with file times being updated incorrectly
> so I am a little sensitive this aspect.
> 
> >>> +int set_page_dirty_mapping(struct page *page);
> >>>   
> >>>       
> >> This aspect of the design seems intrusive to me.  I didn't see a strong
> >> reason to introduce new versions of many of the routines just to handle
> >> these semantics.  What motivated this part of your design?  Why the new
> >> _mapping versions of routines?
> >>     
> >
> > Because there's no way to know inside the set_page_dirty() functions
> > if the dirtying comes from a memory mapping or from a modification
> > through a normal write().  And they have different semantics, for
> > write() the modification times are updated immediately.
> 
> Perhaps I didn't understand what page_mapped() does, but it does seem to
> have the right semantics as far as I could see.

The problems will start, when you have a file that is both mapped and
modified with write().  Then the dirying from the write() will set the
flag, and that will have undesirable consequences.

Miklos
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/