Skip to content
Snippets Groups Projects
Select Git revision
  • 0604475119de5f80dc051a5db055c6a2a75bd542
  • vme-testing default
  • ci-test
  • master
  • remoteproc
  • am625-sk-ov5640
  • pcal6534-upstreaming
  • lps22df-upstreaming
  • msc-upstreaming
  • imx8mp
  • iio/noa1305
  • vme-next
  • vme-next-4.14-rc4
  • v4.14-rc4
  • v4.14-rc3
  • v4.14-rc2
  • v4.14-rc1
  • v4.13
  • vme-next-4.13-rc7
  • v4.13-rc7
  • v4.13-rc6
  • v4.13-rc5
  • v4.13-rc4
  • v4.13-rc3
  • v4.13-rc2
  • v4.13-rc1
  • v4.12
  • v4.12-rc7
  • v4.12-rc6
  • v4.12-rc5
  • v4.12-rc4
  • v4.12-rc3
32 results

tcp_ipv6.c

Blame
    • Eric Dumazet's avatar
      06044751
      tcp: add TCPMemoryPressuresChrono counter · 06044751
      Eric Dumazet authored
      
      DRAM supply shortage and poor memory pressure tracking in TCP
      stack makes any change in SO_SNDBUF/SO_RCVBUF (or equivalent autotuning
      limits) and tcp_mem[] quite hazardous.
      
      TCPMemoryPressures SNMP counter is an indication of tcp_mem sysctl
      limits being hit, but only tracking number of transitions.
      
      If TCP stack behavior under stress was perfect :
      1) It would maintain memory usage close to the limit.
      2) Memory pressure state would be entered for short times.
      
      We certainly prefer 100 events lasting 10ms compared to one event
      lasting 200 seconds.
      
      This patch adds a new SNMP counter tracking cumulative duration of
      memory pressure events, given in ms units.
      
      $ cat /proc/sys/net/ipv4/tcp_mem
      3088    4117    6176
      $ grep TCP /proc/net/sockstat
      TCP: inuse 180 orphan 0 tw 2 alloc 234 mem 4140
      $ nstat -n ; sleep 10 ; nstat |grep Pressure
      TcpExtTCPMemoryPressures        1700
      TcpExtTCPMemoryPressuresChrono  5209
      
      v2: Used EXPORT_SYMBOL_GPL() instead of EXPORT_SYMBOL() as David
      instructed.
      
      Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
      06044751
      History
      tcp: add TCPMemoryPressuresChrono counter
      Eric Dumazet authored
      
      DRAM supply shortage and poor memory pressure tracking in TCP
      stack makes any change in SO_SNDBUF/SO_RCVBUF (or equivalent autotuning
      limits) and tcp_mem[] quite hazardous.
      
      TCPMemoryPressures SNMP counter is an indication of tcp_mem sysctl
      limits being hit, but only tracking number of transitions.
      
      If TCP stack behavior under stress was perfect :
      1) It would maintain memory usage close to the limit.
      2) Memory pressure state would be entered for short times.
      
      We certainly prefer 100 events lasting 10ms compared to one event
      lasting 200 seconds.
      
      This patch adds a new SNMP counter tracking cumulative duration of
      memory pressure events, given in ms units.
      
      $ cat /proc/sys/net/ipv4/tcp_mem
      3088    4117    6176
      $ grep TCP /proc/net/sockstat
      TCP: inuse 180 orphan 0 tw 2 alloc 234 mem 4140
      $ nstat -n ; sleep 10 ; nstat |grep Pressure
      TcpExtTCPMemoryPressures        1700
      TcpExtTCPMemoryPressuresChrono  5209
      
      v2: Used EXPORT_SYMBOL_GPL() instead of EXPORT_SYMBOL() as David
      instructed.
      
      Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
      Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
    mxs_timer.c 7.32 KiB
    // SPDX-License-Identifier: GPL-2.0+
    //
    //  Copyright (C) 2000-2001 Deep Blue Solutions
    //  Copyright (C) 2002 Shane Nay (shane@minirl.com)
    //  Copyright (C) 2006-2007 Pavel Pisa (ppisa@pikron.com)
    //  Copyright (C) 2008 Juergen Beisert (kernel@pengutronix.de)
    //  Copyright (C) 2010 Freescale Semiconductor, Inc. All Rights Reserved.
    
    #include <linux/err.h>
    #include <linux/interrupt.h>
    #include <linux/irq.h>
    #include <linux/clockchips.h>
    #include <linux/clk.h>
    #include <linux/of.h>
    #include <linux/of_address.h>
    #include <linux/of_irq.h>
    #include <linux/stmp_device.h>
    #include <linux/sched_clock.h>
    
    /*
     * There are 2 versions of the timrot on Freescale MXS-based SoCs.
     * The v1 on MX23 only gets 16 bits counter, while v2 on MX28
     * extends the counter to 32 bits.
     *
     * The implementation uses two timers, one for clock_event and
     * another for clocksource. MX28 uses timrot 0 and 1, while MX23
     * uses 0 and 2.
     */
    
    #define MX23_TIMROT_VERSION_OFFSET	0x0a0
    #define MX28_TIMROT_VERSION_OFFSET	0x120
    #define BP_TIMROT_MAJOR_VERSION		24
    #define BV_TIMROT_VERSION_1		0x01
    #define BV_TIMROT_VERSION_2		0x02
    #define timrot_is_v1()	(timrot_major_version == BV_TIMROT_VERSION_1)
    
    /*
     * There are 4 registers for each timrotv2 instance, and 2 registers
     * for each timrotv1. So address step 0x40 in macros below strides
     * one instance of timrotv2 while two instances of timrotv1.
     *
     * As the result, HW_TIMROT_XXXn(1) defines the address of timrot1
     * on MX28 while timrot2 on MX23.
     */
    /* common between v1 and v2 */
    #define HW_TIMROT_ROTCTRL		0x00
    #define HW_TIMROT_TIMCTRLn(n)		(0x20 + (n) * 0x40)
    /* v1 only */
    #define HW_TIMROT_TIMCOUNTn(n)		(0x30 + (n) * 0x40)
    /* v2 only */
    #define HW_TIMROT_RUNNING_COUNTn(n)	(0x30 + (n) * 0x40)
    #define HW_TIMROT_FIXED_COUNTn(n)	(0x40 + (n) * 0x40)
    
    #define BM_TIMROT_TIMCTRLn_RELOAD	(1 << 6)
    #define BM_TIMROT_TIMCTRLn_UPDATE	(1 << 7)
    #define BM_TIMROT_TIMCTRLn_IRQ_EN	(1 << 14)
    #define BM_TIMROT_TIMCTRLn_IRQ		(1 << 15)
    #define BP_TIMROT_TIMCTRLn_SELECT	0
    #define BV_TIMROTv1_TIMCTRLn_SELECT__32KHZ_XTAL		0x8
    #define BV_TIMROTv2_TIMCTRLn_SELECT__32KHZ_XTAL		0xb
    #define BV_TIMROTv2_TIMCTRLn_SELECT__TICK_ALWAYS	0xf
    
    static struct clock_event_device mxs_clockevent_device;
    
    static void __iomem *mxs_timrot_base;
    static u32 timrot_major_version;
    
    static inline void timrot_irq_disable(void)
    {
    	__raw_writel(BM_TIMROT_TIMCTRLn_IRQ_EN, mxs_timrot_base +
    		     HW_TIMROT_TIMCTRLn(0) + STMP_OFFSET_REG_CLR);
    }
    
    static inline void timrot_irq_enable(void)
    {
    	__raw_writel(BM_TIMROT_TIMCTRLn_IRQ_EN, mxs_timrot_base +
    		     HW_TIMROT_TIMCTRLn(0) + STMP_OFFSET_REG_SET);
    }
    
    static void timrot_irq_acknowledge(void)
    {
    	__raw_writel(BM_TIMROT_TIMCTRLn_IRQ, mxs_timrot_base +
    		     HW_TIMROT_TIMCTRLn(0) + STMP_OFFSET_REG_CLR);
    }
    
    static u64 timrotv1_get_cycles(struct clocksource *cs)
    {
    	return ~((__raw_readl(mxs_timrot_base + HW_TIMROT_TIMCOUNTn(1))
    			& 0xffff0000) >> 16);
    }
    
    static int timrotv1_set_next_event(unsigned long evt,
    					struct clock_event_device *dev)
    {
    	/* timrot decrements the count */
    	__raw_writel(evt, mxs_timrot_base + HW_TIMROT_TIMCOUNTn(0));
    
    	return 0;
    }
    
    static int timrotv2_set_next_event(unsigned long evt,
    					struct clock_event_device *dev)
    {
    	/* timrot decrements the count */
    	__raw_writel(evt, mxs_timrot_base + HW_TIMROT_FIXED_COUNTn(0));
    
    	return 0;
    }
    
    static irqreturn_t mxs_timer_interrupt(int irq, void *dev_id)
    {
    	struct clock_event_device *evt = dev_id;
    
    	timrot_irq_acknowledge();
    	evt->event_handler(evt);
    
    	return IRQ_HANDLED;
    }
    
    static void mxs_irq_clear(char *state)
    {
    	/* Disable interrupt in timer module */
    	timrot_irq_disable();
    
    	/* Set event time into the furthest future */
    	if (timrot_is_v1())
    		__raw_writel(0xffff, mxs_timrot_base + HW_TIMROT_TIMCOUNTn(1));
    	else
    		__raw_writel(0xffffffff,
    			     mxs_timrot_base + HW_TIMROT_FIXED_COUNTn(1));
    
    	/* Clear pending interrupt */
    	timrot_irq_acknowledge();
    	pr_debug("%s: changing mode to %s\n", __func__, state);
    }
    
    static int mxs_shutdown(struct clock_event_device *evt)
    {
    	mxs_irq_clear("shutdown");
    
    	return 0;
    }
    
    static int mxs_set_oneshot(struct clock_event_device *evt)
    {
    	if (clockevent_state_oneshot(evt))
    		mxs_irq_clear("oneshot");
    	timrot_irq_enable();
    	return 0;
    }
    
    static struct clock_event_device mxs_clockevent_device = {
    	.name			= "mxs_timrot",
    	.features		= CLOCK_EVT_FEAT_ONESHOT,
    	.set_state_shutdown	= mxs_shutdown,
    	.set_state_oneshot	= mxs_set_oneshot,
    	.tick_resume		= mxs_shutdown,
    	.set_next_event		= timrotv2_set_next_event,
    	.rating			= 200,
    };
    
    static int __init mxs_clockevent_init(struct clk *timer_clk)
    {
    	if (timrot_is_v1())
    		mxs_clockevent_device.set_next_event = timrotv1_set_next_event;
    	mxs_clockevent_device.cpumask = cpumask_of(0);
    	clockevents_config_and_register(&mxs_clockevent_device,
    					clk_get_rate(timer_clk),
    					timrot_is_v1() ? 0xf : 0x2,
    					timrot_is_v1() ? 0xfffe : 0xfffffffe);
    
    	return 0;
    }
    
    static struct clocksource clocksource_mxs = {
    	.name		= "mxs_timer",
    	.rating		= 200,
    	.read		= timrotv1_get_cycles,
    	.mask		= CLOCKSOURCE_MASK(16),
    	.flags		= CLOCK_SOURCE_IS_CONTINUOUS,
    };
    
    static u64 notrace mxs_read_sched_clock_v2(void)
    {
    	return ~readl_relaxed(mxs_timrot_base + HW_TIMROT_RUNNING_COUNTn(1));
    }
    
    static int __init mxs_clocksource_init(struct clk *timer_clk)
    {
    	unsigned int c = clk_get_rate(timer_clk);
    
    	if (timrot_is_v1())
    		clocksource_register_hz(&clocksource_mxs, c);
    	else {
    		clocksource_mmio_init(mxs_timrot_base + HW_TIMROT_RUNNING_COUNTn(1),
    			"mxs_timer", c, 200, 32, clocksource_mmio_readl_down);
    		sched_clock_register(mxs_read_sched_clock_v2, 32, c);
    	}
    
    	return 0;
    }
    
    static int __init mxs_timer_init(struct device_node *np)
    {
    	struct clk *timer_clk;
    	int irq, ret;
    
    	mxs_timrot_base = of_iomap(np, 0);
    	WARN_ON(!mxs_timrot_base);
    
    	timer_clk = of_clk_get(np, 0);
    	if (IS_ERR(timer_clk)) {
    		pr_err("%s: failed to get clk\n", __func__);
    		return PTR_ERR(timer_clk);
    	}
    
    	ret = clk_prepare_enable(timer_clk);
    	if (ret)
    		return ret;
    
    	/*
    	 * Initialize timers to a known state
    	 */
    	stmp_reset_block(mxs_timrot_base + HW_TIMROT_ROTCTRL);
    
    	/* get timrot version */
    	timrot_major_version = __raw_readl(mxs_timrot_base +
    			(of_device_is_compatible(np, "fsl,imx23-timrot") ?
    						MX23_TIMROT_VERSION_OFFSET :
    						MX28_TIMROT_VERSION_OFFSET));
    	timrot_major_version >>= BP_TIMROT_MAJOR_VERSION;
    
    	/* one for clock_event */
    	__raw_writel((timrot_is_v1() ?
    			BV_TIMROTv1_TIMCTRLn_SELECT__32KHZ_XTAL :
    			BV_TIMROTv2_TIMCTRLn_SELECT__TICK_ALWAYS) |
    			BM_TIMROT_TIMCTRLn_UPDATE |
    			BM_TIMROT_TIMCTRLn_IRQ_EN,
    			mxs_timrot_base + HW_TIMROT_TIMCTRLn(0));
    
    	/* another for clocksource */
    	__raw_writel((timrot_is_v1() ?
    			BV_TIMROTv1_TIMCTRLn_SELECT__32KHZ_XTAL :
    			BV_TIMROTv2_TIMCTRLn_SELECT__TICK_ALWAYS) |
    			BM_TIMROT_TIMCTRLn_RELOAD,
    			mxs_timrot_base + HW_TIMROT_TIMCTRLn(1));
    
    	/* set clocksource timer fixed count to the maximum */
    	if (timrot_is_v1())
    		__raw_writel(0xffff,
    			mxs_timrot_base + HW_TIMROT_TIMCOUNTn(1));
    	else
    		__raw_writel(0xffffffff,
    			mxs_timrot_base + HW_TIMROT_FIXED_COUNTn(1));
    
    	/* init and register the timer to the framework */
    	ret = mxs_clocksource_init(timer_clk);
    	if (ret)
    		return ret;
    
    	ret = mxs_clockevent_init(timer_clk);
    	if (ret)
    		return ret;
    
    	/* Make irqs happen */
    	irq = irq_of_parse_and_map(np, 0);
    	if (irq <= 0)
    		return -EINVAL;
    
    	return request_irq(irq, mxs_timer_interrupt, IRQF_TIMER | IRQF_IRQPOLL,
    			   "MXS Timer Tick", &mxs_clockevent_device);
    }
    TIMER_OF_DECLARE(mxs, "fsl,timrot", mxs_timer_init);