netdev - Re: [PATCH net-next 3/7] gve: Do lazy cleanup in TX path

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] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 7 Oct 2021 12:57:11 -0700
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Jeroen de Borst <jeroendb@...gle.com>, netdev@...r.kernel.org
Cc:     davem@...emloft.net, kuba@...nel.org,
        Tao Liu <xliutaox@...gle.com>,
        Catherine Sullivan <csully@...gle.com>
Subject: Re: [PATCH net-next 3/7] gve: Do lazy cleanup in TX path



On 10/7/21 9:25 AM, Jeroen de Borst wrote:
> From: Tao Liu <xliutaox@...gle.com>
> 
> When TX queue is full, attemt to process enough TX completions
> to avoid stalling the queue.
> 
> Fixes: f5cedc84a30d2 ("gve: Add transmit and receive support")
> Signed-off-by: Tao Liu <xliutaox@...gle.com>
> Signed-off-by: Catherine Sullivan <csully@...gle.com>
> ---
>  drivers/net/ethernet/google/gve/gve.h         |  9 +-
>  drivers/net/ethernet/google/gve/gve_ethtool.c |  3 +-
>  drivers/net/ethernet/google/gve/gve_main.c    |  6 +-
>  drivers/net/ethernet/google/gve/gve_tx.c      | 94 +++++++++++--------
>  4 files changed, 62 insertions(+), 50 deletions(-)
> 
> diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
> index 59c525800e5d..003b30b91c6d 100644
> --- a/drivers/net/ethernet/google/gve/gve.h
> +++ b/drivers/net/ethernet/google/gve/gve.h
> @@ -341,8 +341,8 @@ struct gve_tx_ring {
>  	union {
>  		/* GQI fields */
>  		struct {
> -			/* NIC tail pointer */
> -			__be32 last_nic_done;
> +			/* Spinlock for when cleanup in progress */
> +			spinlock_t clean_lock;
>  		};
>

This is adding yet another spinlock in tx completion path.

Normally, BQL should kick and you should not fill the queue completely.

Something is not right.

tx completion can take a lot of time, it seems odd to block an
innocent thread in ndo_start_xmit().

Please provide more details in your changelog ?