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: <1a8ee542-f309-4106-818c-fe380cf33266@intel.com>
Date:   Thu, 13 Apr 2023 12:24:16 -0700
From:   Jacob Keller <jacob.e.keller@...el.com>
To:     Heiner Kallweit <hkallweit1@...il.com>,
        Jakub Kicinski <kuba@...nel.org>,
        David Miller <davem@...emloft.net>,
        Paolo Abeni <pabeni@...hat.com>,
        "Eric Dumazet" <edumazet@...gle.com>,
        Realtek linux nic maintainers <nic_swsd@...ltek.com>
CC:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH net-next 2/3] r8169: use new macro
 netif_subqueue_maybe_stop in rtl8169_start_xmit



On 4/13/2023 12:15 PM, Heiner Kallweit wrote:
> Use new net core macro netif_subqueue_maybe_stop in the start_xmit path
> to simplify the code. Whilst at it, set the tx queue start threshold to
> twice the stop threshold. Before values were the same, resulting in
> stopping/starting the queue more often than needed.
> 
> Signed-off-by: Heiner Kallweit <hkallweit1@...il.com>
> ---


Reviewed-by: Jacob Keller <jacob.e.keller@...el.com>

>  drivers/net/ethernet/realtek/r8169_main.c | 42 +++++++----------------
>  1 file changed, 13 insertions(+), 29 deletions(-)
> 
> diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
> index 9f8357bbc..3f0b78fd9 100644
> --- a/drivers/net/ethernet/realtek/r8169_main.c
> +++ b/drivers/net/ethernet/realtek/r8169_main.c
> @@ -30,6 +30,7 @@
>  #include <linux/ipv6.h>
>  #include <asm/unaligned.h>
>  #include <net/ip6_checksum.h>
> +#include <net/netdev_queues.h>
>  
>  #include "r8169.h"
>  #include "r8169_firmware.h"
> @@ -68,6 +69,8 @@
>  #define NUM_RX_DESC	256	/* Number of Rx descriptor registers */
>  #define R8169_TX_RING_BYTES	(NUM_TX_DESC * sizeof(struct TxDesc))
>  #define R8169_RX_RING_BYTES	(NUM_RX_DESC * sizeof(struct RxDesc))
> +#define R8169_TX_STOP_THRS	(MAX_SKB_FRAGS + 1)
> +#define R8169_TX_START_THRS	(2 * R8169_TX_STOP_THRS)
>  
>  #define OCP_STD_PHY_BASE	0xa400
>  
> @@ -4162,13 +4165,9 @@ static bool rtl8169_tso_csum_v2(struct rtl8169_private *tp,
>  	return true;
>  }
>  
> -static bool rtl_tx_slots_avail(struct rtl8169_private *tp)
> +static unsigned int rtl_tx_slots_avail(struct rtl8169_private *tp)
>  {
> -	unsigned int slots_avail = READ_ONCE(tp->dirty_tx) + NUM_TX_DESC
> -					- READ_ONCE(tp->cur_tx);
> -
> -	/* A skbuff with nr_frags needs nr_frags+1 entries in the tx queue */
> -	return slots_avail > MAX_SKB_FRAGS;
> +	return READ_ONCE(tp->dirty_tx) + NUM_TX_DESC - READ_ONCE(tp->cur_tx);
>  }
>  

Ok so now we just directly return the slots available. We used to check
MAX_SKB_FRAGS vs now we check MAX_SBK_FAGS + 1 (or double that for the
start threshhold). Ok.

>  /* Versions RTL8102e and from RTL8168c onwards support csum_v2 */
> @@ -4198,7 +4197,8 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
>  	struct rtl8169_private *tp = netdev_priv(dev);
>  	unsigned int entry = tp->cur_tx % NUM_TX_DESC;
>  	struct TxDesc *txd_first, *txd_last;
> -	bool stop_queue, door_bell;
> +	bool door_bell;
> +	int stop_queue;
>  	u32 opts[2];
>  
>  	if (unlikely(!rtl_tx_slots_avail(tp))) {
> @@ -4245,27 +4245,10 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
>  
>  	WRITE_ONCE(tp->cur_tx, tp->cur_tx + frags + 1);
>  
> -	stop_queue = !rtl_tx_slots_avail(tp);
> -	if (unlikely(stop_queue)) {
> -		/* Avoid wrongly optimistic queue wake-up: rtl_tx thread must
> -		 * not miss a ring update when it notices a stopped queue.
> -		 */
> -		smp_wmb();
> -		netif_stop_queue(dev);
> -		/* Sync with rtl_tx:
> -		 * - publish queue status and cur_tx ring index (write barrier)
> -		 * - refresh dirty_tx ring index (read barrier).
> -		 * May the current thread have a pessimistic view of the ring
> -		 * status and forget to wake up queue, a racing rtl_tx thread
> -		 * can't.
> -		 */
> -		smp_mb__after_atomic();
> -		if (rtl_tx_slots_avail(tp))
> -			netif_start_queue(dev);
> -		door_bell = true;
> -	}
> -
> -	if (door_bell)
> +	stop_queue = netif_subqueue_maybe_stop(dev, 0, rtl_tx_slots_avail(tp),
> +					       R8169_TX_STOP_THRS,
> +					       R8169_TX_START_THRS);
> +	if (door_bell || stop_queue < 0)


Nice reduction in code here :D

>  		rtl8169_doorbell(tp);
>  
>  	return NETDEV_TX_OK;
> @@ -4400,7 +4383,8 @@ static void rtl_tx(struct net_device *dev, struct rtl8169_private *tp,
>  		 * ring status.
>  		 */
>  		smp_store_mb(tp->dirty_tx, dirty_tx);
> -		if (netif_queue_stopped(dev) && rtl_tx_slots_avail(tp))
> +		if (netif_queue_stopped(dev) &&
> +		    rtl_tx_slots_avail(tp) >= R8169_TX_START_THRS)
>  			netif_wake_queue(dev);
>  		/*
>  		 * 8168 hack: TxPoll requests are lost when the Tx packets are

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ