netdev - Re: [PATCH] net: stmmac: fix ethtool per-queue statistics

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]

Message-ID: <20240105192440.25a7b954@meshulam.tesarici.cz>
Date: Fri, 5 Jan 2024 19:24:40 +0100
From: Petr Tesařík <petr@...arici.cz>
To: Alexandre Torgue <alexandre.torgue@...s.st.com>, Jose Abreu
 <joabreu@...opsys.com>, "David S. Miller" <davem@...emloft.net>, Eric
 Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo
 Abeni <pabeni@...hat.com>, Maxime Coquelin <mcoquelin.stm32@...il.com>,
 Jisheng Zhang <jszhang@...nel.org>, netdev@...r.kernel.org (open
 list:STMMAC ETHERNET DRIVER), linux-stm32@...md-mailman.stormreply.com
 (moderated list:ARM/STM32 ARCHITECTURE),
 linux-arm-kernel@...ts.infradead.org (moderated list:ARM/STM32
 ARCHITECTURE), linux-kernel@...r.kernel.org (open list)
Subject: Re: [PATCH] net: stmmac: fix ethtool per-queue  statistics

On Fri,  5 Jan 2024 19:10:24 +0100
Petr Tesarik <petr@...arici.cz> wrote:

> Fix per-queue statistics for devices with more than one queue.
> 
> The output data pointer is currently reset in each loop iteration,
> effectively summing all queue statistics in the first four u64 values.
> 
> The summary values are not even labeled correctly. For example, if eth0 has
> 2 queues, ethtool -S eth0 shows:
> 
>      q0_tx_pkt_n: 374 (actually tx_pkt_n over all queues)
>      q0_tx_irq_n: 23  (actually tx_normal_irq_n over all queues)
>      q1_tx_pkt_n: 462 (actually rx_pkt_n over all queues)
>      q1_tx_irq_n: 446 (actually rx_normal_irq_n over all queues)
>      q0_rx_pkt_n: 0
>      q0_rx_irq_n: 0
>      q1_rx_pkt_n: 0
>      q1_rx_irq_n: 0
> 
> While touching this code, change the pointer type to u64 and get rid of the
> weird pointer arithmetic.

Just to make sure, this fix has nothing to do with my previous stmmac
fix. It's just a fix for something I noticed while working on the other
patch.

I hope this fix can be merged fast, so I can base my other patch on it.

Petr T

> Signed-off-by: Petr Tesarik <petr@...arici.cz>
> Fixes: 133466c3bbe1 ("net: stmmac: use per-queue 64 bit statistics where necessary")
> ---
>  .../ethernet/stmicro/stmmac/stmmac_ethtool.c  | 23 ++++++-------------
>  1 file changed, 7 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
> index f628411ae4ae..023876fc4da7 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
> @@ -543,43 +543,34 @@ static void stmmac_get_per_qstats(struct stmmac_priv *priv, u64 *data)
>  	u32 rx_cnt = priv->plat->rx_queues_to_use;
>  	unsigned int start;
>  	int q, stat;
> -	u64 *pos;
> -	char *p;
> +	u64 *p;
>  
> -	pos = data;
>  	for (q = 0; q < tx_cnt; q++) {
>  		struct stmmac_txq_stats *txq_stats = &priv->xstats.txq_stats[q];
>  		struct stmmac_txq_stats snapshot;
>  
> -		data = pos;
>  		do {
>  			start = u64_stats_fetch_begin(&txq_stats->syncp);
>  			snapshot = *txq_stats;
>  		} while (u64_stats_fetch_retry(&txq_stats->syncp, start));
>  
> -		p = (char *)&snapshot + offsetof(struct stmmac_txq_stats, tx_pkt_n);
> -		for (stat = 0; stat < STMMAC_TXQ_STATS; stat++) {
> -			*data++ += (*(u64 *)p);
> -			p += sizeof(u64);
> -		}
> +		p = &snapshot.tx_pkt_n;
> +		for (stat = 0; stat < STMMAC_TXQ_STATS; stat++)
> +			*data++ = *p++;
>  	}
>  
> -	pos = data;
>  	for (q = 0; q < rx_cnt; q++) {
>  		struct stmmac_rxq_stats *rxq_stats = &priv->xstats.rxq_stats[q];
>  		struct stmmac_rxq_stats snapshot;
>  
> -		data = pos;
>  		do {
>  			start = u64_stats_fetch_begin(&rxq_stats->syncp);
>  			snapshot = *rxq_stats;
>  		} while (u64_stats_fetch_retry(&rxq_stats->syncp, start));
>  
> -		p = (char *)&snapshot + offsetof(struct stmmac_rxq_stats, rx_pkt_n);
> -		for (stat = 0; stat < STMMAC_RXQ_STATS; stat++) {
> -			*data++ += (*(u64 *)p);
> -			p += sizeof(u64);
> -		}
> +		p = &snapshot.rx_pkt_n;
> +		for (stat = 0; stat < STMMAC_RXQ_STATS; stat++)
> +			*data++ = *p++;
>  	}
>  }
>