[OpenAFS-devel] Re: progress... sortof...

[Nickolai Zeldovich]

On Wed, 2004-04-28 at 20:53 -0500, Neulinger, Nathan wrote:
> It's not all the time... but certain user written code definately
> triggers the problem, and once it's hit once, it takes the machine down.
> I hae one set of user code (unfortunately not very minimal) that seems
> to trigger it very reliably on some of the boxes I have. 

Hm..  It'd be curious if you could instrument osi_file_uio_rdwr and see
when you get EOF reading a file, and also instrument afs_UFSRead to see
how the invalid length is being calculated?  Maybe we don't have enough
locking around files and there's a race condition so the file is being
truncated...

> Shouldn't that be -EIO?

Doesn't seem like it..  We explicitly make error codes positive in that
function, when FOP_READ returns a negative value.

-- kolya