On Tue, Mar 19, 2024 at 11:06:01PM +0100, Christoph Hellwig wrote:
On Tue, Mar 19, 2024 at 08:47:10PM +0000, Chuck Lever III wrote:
That comment doesn't match the size and complexity of
the three patches I applied. They are tiny, adding one
internal block device method call. I don't see any
palpable risk here, please elaborate.
Anything touching the interface between subsystems is sketchy
and potentially dangerous.
Do we have any tests for this functionality? If we do, we can ask Chuck
to run them. If we don't, how did it land upstream?
And I'm rally done with this BS discussion here. There is obviously
nothing fixed by backporting it, there has been no testing on the
backport and we're all wasting our time.
There is nothing fixed by these particular patches, but it keeps us
aligned with upstream and makes further backports easier.
We've seen so many issues with custom backports that diverge from
upstream to the point where we'd rather take new functionality than deal
with those (and then manually resolve conflicts for the next few years
in this subsystem).
NAK for these backports.
What's the alternative? Would you sign up for backporting future patches
that are affected by this divergence?
--
Thanks,
Sasha
