netdev - Re: [PATCH net-next] Revert "crypto: chelsio

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>] [<thread-prev] [day] [month] [year] [list]

Message-Id: <20200506.145220.1567628861970386108.davem@davemloft.net>
Date:   Wed, 06 May 2020 14:52:20 -0700 (PDT)
From:   David Miller <davem@...emloft.net>
To:     ayush.sawal@...lsio.com
Cc:     herbert@...dor.apana.org.au, linux-crypto@...r.kernel.org,
        netdev@...r.kernel.org, manojmalaviya@...lsio.com
Subject: Re: [PATCH net-next] Revert "crypto: chelsio - Inline single pdu
 only"

From: Ayush Sawal <ayush.sawal@...lsio.com>
Date: Wed,  6 May 2020 20:17:19 +0530

> This reverts commit 27c6feb0fb33a665a746346e76714826a5be5d10.
> 
> For ipsec offload the chelsio's ethernet driver expects a single mtu
> sized packet.
> 
> But when ipsec traffic is running using iperf, most of the packets in
> that traffic are gso packets(large sized skbs) because GSO is enabled by
> default in TCP, due to this commit 0a6b2a1dc2a2 ("tcp: switch to GSO
> being always on"), so chcr_ipsec_offload_ok() receives a gso
> skb(with gso_size non zero).
> 
> Due to the check in chcr_ipsec_offload_ok(), this function returns false
> for most of the packet, then ipsec offload is skipped and the skb goes
> out taking the coprocessor path which reduces the bandwidth for inline
> ipsec.
> 
> If this check is removed then for most of the packets(large sized skbs)
> the chcr_ipsec_offload_ok() returns true and then as GSO is on, the
> segmentation of the packet happens in the kernel and then finally the
> driver_xmit is called, which receives a segmented mtu sized packet which
> is what the driver expects for ipsec offload. So this case becomes
> unnecessary here, therefore removing it.
> 
> Signed-off-by: Ayush Sawal <ayush.sawal@...lsio.com>

Applied.