On Oct 4, 2024, at 1:59 PM, Youzhong Yang youzhong@gmail.com wrote:
The explanation of how it can happen is in the commit message. Using list_head 'nf_lru' for two purposes (the LRU list and dispose list) is problematic.
"is problematic" is not an adequate or precise explanation of how the code is failing. But as I said, I'm not objecting, just noting that we don't understand why this change addresses the problem.
In other words, we have test experience that shows the patch is safe to apply, but no deep explanation of why it is effective.
I also mentioned my reproducer in one of the e-mail threads, here it is if it still matters:
I'm asking that you please apply and test these patches on v6.11 and v6.1, at the very least, before requesting that Greg apply these patches to the LTS kernels. Greg needs very clear instructions about how he should proceed, as well as some evidence that we are not asking him to apply patches that will break the target kernels.
And, please confirm that 4/4 is needed. I can't see any obvious reason why it is necessary to prevent a leak.
-- Chuck Lever