Re: drivers/char/random.c: more ruminations

["George Spelvin" <linux@horizon.com>]

> So have you actually instrumented the kernel to demonstrate that in
> fact we have super deep stack call paths where the 128 bytes worth of
> stack actually matters?

I haven't got a specific call chain where 128 bytes pushes it
over a limit.  But kernel stack usage is a perennial problem.
Wasn't there some discussion about that just recenty?
6538b8ea8: "x86_64: expand kernel stack to 16K"

I agree a 128 byte stack frame is not one of the worst offenders,
but it's enough to try to clean up if possible.

You can search LKML for a bunch of discussion of 176 bytes
in __alloc_pages_slowpath().

And in this case, it's so *easy*.  extract_buf() works 10 bytes at a
time anyway, and _mix_pool_bytes is byte at a time.

>> I hadn't tested the patch when I mailed it to you (I prepared it in
>> order to reply to your e-mail, and it's annoying to reboot the machine
>> I'm composing an e-mail on), but I have since.  It works.

> As an aside, I'd strongly suggest that you use kvm to do your kernel
> testing.  It means you can do a lot more testing which is always a
> good thing....

H'mmm. I need to learn what KVM *is*.  Apparently there's a second
meaning other than "keyboard, video & mouse". :-)

Normally, I just test using modules.  Especially when working on a
driver for a hardware device, virtualization makes life difficult.
But /dev/random is (for good reasons) not modularizable.

(I can see how it'd be useful for filesystem development, however.)