On Mon, 10 Feb 2025 10:51:31 +0100 Stefano Brivio <sbrivio(a)redhat.com> wrote:On Mon, 10 Feb 2025 17:05:08 +1100 David Gibson <david(a)gibson.dropbear.id.au> wrote:Oh, I missed the fact you already split this out in v14... fine then. There are some stray FIXMEs in 11/14, tcp_flow_migrate_source_ext(): + /* FIXME: Reenable dump in tcp_flow_migrate_shrink_window() */ + /* FIXME: it's either this or flow_migrate_source_early(), why? */ but I'll just drop them in the next version. -- StefanoOn Sun, Feb 09, 2025 at 11:20:04PM +0100, Stefano Brivio wrote:I'd rather not because, due to this, the code for the case *without it* is a bit different anyway, and we avoid some code churn. I would also like to merge it unused to make our lives a bit easier the day we retry to work on it.+++ b/migrate.c @@ -98,11 +98,30 @@ static int seen_addrs_target_v1(struct ctx *c, /* Stages for version 1 */ static const struct migrate_stage stages_v1[] = { + /* FIXME: With this step, close() in tcp_flow_migrate_source_ext() + * *sometimes* closes the connection for real. + */ +/* { + .name = "shrink TCP windows", + .source = flow_migrate_source_early, + .target = NULL, + }, +*/Given we're not sure if this will help, and it adds some complications, probably makes sense to split this into a separate patch.