Driver Update Disk discussion

[Peter Jeremy]

On Thu, 2005-Feb-24 17:59:19 -0700, Scott Long wrote:
>- kernel option support.  How do we support vendor modules in a kernel
>that might be compiled with PAE (rather common these days), SMP, MAC,
>etc.  The loader and /boot infrastructure has no concept of this.  It's
>highly important, though.

AFAIK, PAE is only relevant on iA32.  I second the suggestion that PAE
be treated as a distinct architecture for these purposes.

INVARIANTS and WITNESS are the other options that impact ABI.  These
are probably unnecessary on -RELEASE but it would be nice if people
could build a kernel with WITNESS and not have it panic if they loaded
a module that wasn't compiled with WITNESS (which I think it the
current behaviour).

>- kernel api/abi.  We are trying to keep the kernel api/abi stable now,
...
>don't have the right hash.  Should we follow with something similar, or
>should we have runtime checks that check symbol/structure signatures?

It would be wonderful if we could have a mechanism that did load-time
validation that the module was compatible with the kernel.  Unfortunately,
I don't think there's any sane way to verify data structure compatability.
(Verifying function call APIs is reasonably easy).  Run-time checking adds
overheads which may be significant for commonly used interfaces.

>Or should we say that we make no guarantees about a binary-only module
>working on anything but a -RELEASE kernel?

At the very least we need to support errata branches.  The RE team has
expended a lot of effort to provide a mechanism to handle critical
problems found post-release.  We don't want to negate this by telling
users that they have a choice of either using driver X or fixing hole Y.

Unfortunately, in the rare case where an errata fix affects a kernel
API/ABI, the change is probably critical to fixing the problem.  This
would require the FreeBSD fix to be co-ordinated with the driver vendor.

I think that guaranteeing operation in -STABLE is probably impractical -
though API/ABI changes would need to be flagged to vendors so they could
test drivers for the next FreeBSD release.

-- 
Peter Jeremy