[Info-vax] Assembly languages

[Simon Clubley]

On 2022-04-12, VAXman-  @SendSpamHere.ORG <VAXman- at SendSpamHere.ORG> wrote:
> In article <t33r9h$pr$1 at dont-email.me>, Simon Clubley <clubley at remove_me.eisner.decus.org-Earth.UFP> writes:
>>On 2022-04-11, VAXman-  @SendSpamHere.ORG <VAXman- at SendSpamHere.ORG> wrote:
>>> In article <t31ose$pr0$1 at dont-email.me>, Simon Clubley <clubley at remove_me.eisner.decus.org-Earth.UFP> writes:
>>>>
>>>>Ok, Brian, you win. I'll be pedantic if you wish. :-)
>>>>
>>>>Once you have code you control running in one of the hardware inner modes,
>>>>you can get to the others without any additional privileges required on
>>>>the part of the account doing it.
>>>
>>> NOT TRUE.  Stop confusing $CMKRNL from EXEC mode with all others.  You can 
>>> NOT get to EXEC from SUPERVISOR mode.  Granted, you found an exploit with an
>>> installed image but that was corrected.  There's no $CMEXEC jump from SUPER-
>>> VISOR mode without privileges vis-a-vis $CMKRNL from EXEC mode.
>>>
>>
>>You are contradicting yourself with the above statements Brian.
>>First you say it's not possible, then you say it's possible if the
>>supervisor mode code has access to the privileges of the current image.
>
> IT IS NOT THAT SUPERVISOR MODE GRANTED YOU ANY SPECIAL MOVEMENT TO AN INNER
> MODE!  You used an image installed with privileges.  The gang that found the
> SMG$ bug employed an image installed with privileges FROM USER MODE.  Bugs
> occur in software save, of course, yours -- assuming you can write software
> or do you just spend your day littering c.o.v? -- because you are perfect in
> every way and noone can question your assertions and aspersions.
>

In a way, it did. The image activator behaves differently when invoked
in supervisor mode compared to when invoked in user mode.

The key question is this: Can a non-privileged user who gets code they
control running in supervisor mode come up with a way to switch to
from supervisor mode to executive mode or kernel mode ?

The answer to that question is yes.

Which means that my original statement about VMS only having a single
inner mode when looked at from the viewpoint of security is also correct.

Simon.

PS: BTW, I don't assume that I am perfect when I write code. That's why
I like languages that help me find errors early on.

-- 
Simon Clubley, clubley at remove_me.eisner.decus.org-Earth.UFP
Walking destinations on a map are further away than they appear.