Re: Tagged memory (was Re: Memory Upgrade)

you are singing IBM's party tune. Wake up to the real world.

----- Original Message -----
From: Randy Mangham <randy.mangham@home.com>
To: <midrange-l@midrange.com>
Sent: Thursday, September 27, 2001 9:30 PM
Subject: Re: Tagged memory (was Re: Memory Upgrade)


> With regard to tags and whether or not the tag bit can be fiddled with thus
> violating system security, I think Frank Soltis probably has the last and
> authoritative word on the subject. I quote from Inside the AS/400, 2nd ed.
> The capital letter emphasis is my addition.
>
> "The tag instructions are used only by SLIC; they are not generated by the
> translator for MI programs. This means that any store to memory that is
> generated by an MI program used the standard instructions and ALWAYS TURNS
> OFF the tag bits. When a pointer is created as part of a resolve operation,
> SLIC builds the pointer in two 64-bit registers and uses the stq
> instruction to turn the tag bits on in memory. Whenever an MI program
> attempts to use a pointer, SLIC uses the lq instruction to load the
> contents of the pointer into registers and then tests to see whether the
> tag bits are still on. If the tag bits are found to be off, someone
> modified the pointer and, therefore, it is invalid.
>
> Tag bits in the AS/400 do NOT prevent the modification of pointers. Tag
> bits are used to DETECT the modification when the pointer is used. This
> approach is different from the one most memory-protection schemes use.
> Typically, memory protection prevents the modification. This is also
> important, and later we will see that the AS/400 has this type of
> protection on a page basis. For pointers, however, the modification is
> detected after the fact. This approach reduced the amount of hardware
> needed in the early implementations of the system and still provided the
> needed level of protection.
>
> Pointers CANNOT be counterfeited. Tags ensure that a pointer was created by
> the operating system (SLIC) and that it has not been modified by anyone
> other than SLIC. Anyone else who creates a pointer, copies a pointer, or
> modifies a pointer has NO WAY [emphasis added] to turn the tag bits on and
> will end up with a useless 16-byte entity."
>
> I hope that clears up any confusion. I'd highly recommend this book if you
> want a reasonably easy to understand brief explanation of the PowerPC
> processor and how tag bits work overall. The original PowerPC architecture
> did not support tag bits so a special implementation of the chip was used
> in the AS/400 until 1997 that supported a "tags active" mode. From 1997
> forward, the PowerPC chips have supported "tags-active" and "tags-inactive"
> modes and that has allowed the same processor chip to be used in both
> AS/400-iSeries and RS/6000-pSeries machines.
>
> Randy Mangham
> Pacific Crest Consulting
> San Diego CA
>
>
> _______________________________________________
> This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
> To post a message email: MIDRANGE-L@midrange.com
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/cgi-bin/listinfo/midrange-l
> or email: MIDRANGE-L-request@midrange.com
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/midrange-l.
>