Re: [PATCH net-next v5 01/10] tls: Bound consecutive no-data records in tls_sw_read_sock()
flat view
From: Chuck Lever <cel@kernel.org>
Date: 2026-10-19 15:09:07
Also in:
linux-kselftest
On Sun, Oct 18, 2026, at 3:22 PM, netdev-bot+sashiko@kernel.org wrote:
- [High] The new limit only counts consecutive empty records, so a peer can get around it by mixing in small records.
A record that delivers bytes runs the read_actor, and the consumer can stop there. The cap bounds the run during which the read_actor is never called and the consumer cannot act. v6 will make that the argument in the description and drop the sentence about desc->count, which invites the reading above.
- [Medium] When tls_sw_read_sock() stops at TLS_RX_NODATA_LIMIT, it queues data_ready_work only if tls_strp_msg_ready(ctx) is true…
Confirmed, but not specific to the cap. Any parse that strp->work completes on an nvme-tcp socket announces through saved_data_ready, which nvme-tcp never replaced. Before this patch the loop retried tls_strp_check_rcv() once more, so the stall needed two allocation failures instead of one. The announce path needs its own fix; v6 leaves it alone.
- [Low] (Baseline: not confirmed after checking the end of the series.) At this commit, the wake condition at the new TLS_RX_NODATA_LIMIT exit…
Agreed. Patch 5 keeps empty records off rx_list, which closes this at the end of the series. pw-bot: cr -- Chuck Lever (Come to NFS bake-a-thon! https://nfsv4bat.org)