On 04.03.24 11:31, Dmitry Antipov wrote:
Jakub,
could you please check whether an official maintainers of net/smc are
actually active? I'm interesting just because there was no feedback on
[1]. After all, it's still a kernel memory leak, and IMO should not be
silently ignored by the maintainers (if any).
Thanks,
Dmitry
[1]
https://lore.kernel.org/netdev/20240221051608.43241-1-dmantipov@xxxxxxxxx/
Hi Dmitry,
I'm on the way to answering you. I understand your worry and appreciate
your sugguestion on the improvement. Since I'm not the original author,
either, I also need to undestand what was the original intention. i.e.
Why should the fasync_list of the smc socket be handed over to the clc
socket? Is there a way to deal with the list prior to the fallback?
AIU, the syzbot's reports on whichever the original fixed or your last
patch fixed are about the same issue. And both of the fixes seem not to
solve the problem. Instead of patches on patches, I'd prefer to find
the root problem and solve it.
Thus, to the proposed patches from you guys (and back to the question at
the beginning), if the fasyn_list should be handed over, I like the Wen
Gu's patch more. Otherwise, I'd like yours more, but as you already
underlied, it should be done in some other way
Thanks,
Wenjia
