Re: Regression introduced by bfcfaa77bdf0f775263e906015982a608df01c76(vfs: use 'unsigned long' accesses for dcache name comparison and hashing)

[Linus Torvalds]

On Thu, Mar 22, 2012 at 1:36 PM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
>
> Guys, could you check if this fixes it?

Yup, that's the patch I was thinking of too.

But I didn't test it. And considering that apparently I don't have
anything that ever triggers this, I guess I should write some
user-space harness thing for this.

I actually wrote some *other* user-space testing functions (checking
that the hash values we generated were good), but I never tested the
"unimportant and totally obvious" full_name_hash() function.

Btw, about test harnesses: I do have a patch to the kernel that
exposes the dentry hash chains through /proc. Interesting to anybody
else? I decided that I didn't want to try to simulate the dentry
hashing in user space, since part of the final hash function is the
'parent' dentry pointer itself (the name hash part is user-spaceable,
but I wanted to see the distribution of those parent bits too).

                          Linus
--
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/