netdev - Query about tls patch

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <DB7PR04MB42529ADA5780B2E7551329848B570@DB7PR04MB4252.eurprd04.prod.outlook.com>
Date:   Sun, 22 Jul 2018 06:08:47 +0000
From:   Vakul Garg <vakul.garg@....com>
To:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Query about tls patch

Hi 

I got a query reading patch https://patchwork.ozlabs.org/patch/943442/ (already merged).

[PATCH]: tls: Fix zerocopy_from_iter iov handling

In tls_sw_sendmsg(), if zerocopy_from_iter() fails, we go to fallback_to_reg_send.
Here we first call iov_iter_revert(). But the iov_iter_advance  didn't happen in first place.

Similarly, in zerocopy_from_iter(), if 'revert' is passed as 'true', iov_iter_revert() is called even if iov_iter_advance() never happened (case of error return).

In tls_sw_recvmsg() flow, the iov_iter_revert() is always triggered. 
With revert happening, I am missing the point how the case where multiple records are decrypted into user space buffer with one invocation of 'recvmsg()' happens?
It seems that successively dequeued tls records would get decrypted at same location in the output buffer and keep overwriting the previous one.

Regards

Vakul