Re: tree corrupted on disk quota full

From: Linus Torvalds <torvalds@osdl.org>
Date: Jan 11, 2007, 22:32 UTC
Message-ID: <Pine.LNX.4.64.0701111428420.3594@woody.osdl.org>
In-Reply-To: <45A6B5E4.3050007@shadowen.org>

On Thu, 11 Jan 2007, Andy Whitcroft wrote:

Show 7 quoted lines

>
> > So I repeat: either you use "xwrite()" (and handle the partial case), or 
> > you use "write_in_full()" (and the partial case is an *ERROR*). There is 
> > no sane middle ground in between those cases.
> 
> Things should be safe in general with the code as it is as we are
> checking the write length.

NO WE ARE NOT.

I already pointed you to write_buffer(). It used to do the right thing. It doesn't any more. And it doesn't, exactly because you converted it away from a loop that did it right, to doing "write_in_full()" and NOT checking the return value properly.

The thing is, if you support partial writes (ie xwrite()), you need to do it in a loop, and then the correct thing to check for is "zero or error".

Once you don't do a loop (ie "write_in_full()" - the whole _point_ is to not do the loop, after all), you need to either expand that check to "zero or error or partial" (which just makes the code _more_ complex), or you need to make "write_in_full()" just return an error for the partial case.

Which is why I'm harping on this issue: either we use "xwrite()", or we fix "write_in_full()" to return errors on partials. Because the "middle ground", where write_in_full() emulates the partial case of "xwrite()" is actually MORE complex than "xwrite()" itself.

		Linus