Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933137AbXFEP3l (ORCPT ); Tue, 5 Jun 2007 11:29:41 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S933078AbXFEP3D (ORCPT ); Tue, 5 Jun 2007 11:29:03 -0400 Received: from ozlabs.org ([203.10.76.45]:36772 "EHLO ozlabs.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933076AbXFEP3A (ORCPT ); Tue, 5 Jun 2007 11:29:00 -0400 Subject: [PATCH 6/6] lguest use hrtimers From: Rusty Russell To: Andrew Morton Cc: lkml - Kernel Mailing List , James Morris , Ingo Molnar , Thomas Gleixner In-Reply-To: <1181055760.14054.47.camel@localhost.localdomain> References: <1181055308.14054.36.camel@localhost.localdomain> <1181055396.14054.38.camel@localhost.localdomain> <1181055483.14054.40.camel@localhost.localdomain> <1181055606.14054.43.camel@localhost.localdomain> <1181055760.14054.47.camel@localhost.localdomain> Content-Type: text/plain Date: Wed, 06 Jun 2007 01:09:56 +1000 Message-Id: <1181056196.14054.53.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.10.1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 8796 Lines: 282 (BTW Thomas, is the check for delta < minimum actually required in our set_next_event function?) >From jmorris@namei.org Wed May 9 23:37:36 2007 Convert lguest to the hrtimer framework, enabling dynamic ticks and high resolution timers. Signed-off-by: James Morris Signed-off-by: Rusty Russell --- drivers/lguest/core.c | 2 - drivers/lguest/hypercalls.c | 10 +---- drivers/lguest/interrupts_and_traps.c | 34 +++++++++++++++-- drivers/lguest/lg.h | 6 ++- drivers/lguest/lguest.c | 63 +++++++++++++++++++++++++++++++-- drivers/lguest/lguest_user.c | 3 + include/linux/lguest.h | 5 ++ 7 files changed, 106 insertions(+), 17 deletions(-) =================================================================== --- a/drivers/lguest/core.c +++ b/drivers/lguest/core.c @@ -329,7 +329,7 @@ int run_guest(struct lguest *lg, unsigne if (lg->halted) { set_current_state(TASK_INTERRUPTIBLE); - schedule_timeout(1); + schedule(); continue; } =================================================================== --- a/drivers/lguest/hypercalls.c +++ b/drivers/lguest/hypercalls.c @@ -44,13 +44,6 @@ static void do_hcall(struct lguest *lg, else guest_pagetable_flush_user(lg); break; - case LHCALL_TIMER_READ: { - u32 now = jiffies; - mb(); - regs->eax = now - lg->last_timer; - lg->last_timer = now; - break; - } case LHCALL_GET_WALLCLOCK: { struct timespec ts; ktime_get_real_ts(&ts); @@ -84,6 +77,9 @@ static void do_hcall(struct lguest *lg, break; case LHCALL_LOAD_TLS: guest_load_tls(lg, regs->edx); + break; + case LHCALL_SET_CLOCKEVENT: + guest_set_clockevent(lg, regs->edx); break; case LHCALL_TS: lg->ts = regs->edx; =================================================================== --- a/drivers/lguest/interrupts_and_traps.c +++ b/drivers/lguest/interrupts_and_traps.c @@ -72,10 +72,6 @@ void maybe_do_interrupt(struct lguest *l if (!lg->lguest_data) return; - - /* If timer has changed, set timer interrupt. */ - if (jiffies != lg->last_timer) - set_bit(0, lg->irqs_pending); /* Mask out any interrupts they have blocked. */ if (copy_from_user(&blk, lg->lguest_data->blocked_interrupts, @@ -240,3 +236,33 @@ void copy_traps(const struct lguest *lg, else default_idt_entry(&idt[i], i, def[i]); } + +void guest_set_clockevent(struct lguest *lg, unsigned long delta) +{ + ktime_t expires; + + if (unlikely(delta == 0)) { + /* Clock event device is shutting down. */ + hrtimer_cancel(&lg->hrt); + return; + } + + expires = ktime_add_ns(ktime_get_real(), delta); + hrtimer_start(&lg->hrt, expires, HRTIMER_MODE_ABS); +} + +static enum hrtimer_restart clockdev_fn(struct hrtimer *timer) +{ + struct lguest *lg = container_of(timer, struct lguest, hrt); + + set_bit(0, lg->irqs_pending); + if (lg->halted) + wake_up_process(lg->tsk); + return HRTIMER_NORESTART; +} + +void init_clockdev(struct lguest *lg) +{ + hrtimer_init(&lg->hrt, CLOCK_REALTIME, HRTIMER_MODE_ABS); + lg->hrt.function = clockdev_fn; +} =================================================================== --- a/drivers/lguest/lg.h +++ b/drivers/lguest/lg.h @@ -134,7 +134,6 @@ struct lguest u32 cr2; int halted; int ts; - u32 last_timer; u32 next_hcall; u32 esp1; u8 ss1; @@ -173,6 +172,9 @@ struct lguest /* The IDT entries: some copied into lguest_ro_state when running. */ struct desc_struct idt[FIRST_EXTERNAL_VECTOR+LGUEST_IRQS]; struct desc_struct syscall_idt; + + /* Virtual clock device */ + struct hrtimer hrt; /* Pending virtual interrupts */ DECLARE_BITMAP(irqs_pending, LGUEST_IRQS); @@ -202,6 +204,8 @@ void setup_default_idt_entries(struct lg const unsigned long *def); void copy_traps(const struct lguest *lg, struct desc_struct *idt, const unsigned long *def); +void guest_set_clockevent(struct lguest *lg, unsigned long delta); +void init_clockdev(struct lguest *lg); /* segments.c: */ void setup_default_gdt_entries(struct lguest_ro_state *state); =================================================================== --- a/drivers/lguest/lguest.c +++ b/drivers/lguest/lguest.c @@ -26,6 +26,7 @@ #include #include #include +#include #include #include #include @@ -359,8 +360,58 @@ static struct clocksource lguest_clock = .flags = CLOCK_SOURCE_IS_CONTINUOUS, }; +/* We also need a "struct clock_event_device": Linux asks us to set it to go + * off some time in the future. Actually, James Morris figured all this out, I + * just applied the patch. */ +static int lguest_clockevent_set_next_event(unsigned long delta, + struct clock_event_device *evt) +{ + if (delta < LG_CLOCK_MIN_DELTA) { + if (printk_ratelimit()) + printk(KERN_DEBUG "%s: small delta %lu ns\n", + __FUNCTION__, delta); + return -ETIME; + } + hcall(LHCALL_SET_CLOCKEVENT, delta, 0, 0); + return 0; +} + +static void lguest_clockevent_set_mode(enum clock_event_mode mode, + struct clock_event_device *evt) +{ + switch (mode) { + case CLOCK_EVT_MODE_UNUSED: + case CLOCK_EVT_MODE_SHUTDOWN: + /* A 0 argument shuts the clock down. */ + hcall(LHCALL_SET_CLOCKEVENT, 0, 0, 0); + break; + case CLOCK_EVT_MODE_ONESHOT: + /* This is what we expect. */ + break; + case CLOCK_EVT_MODE_PERIODIC: + BUG(); + } +} + +/* This describes our primitive timer chip. */ +static struct clock_event_device lguest_clockevent = { + .name = "lguest", + .features = CLOCK_EVT_FEAT_ONESHOT, + .set_next_event = lguest_clockevent_set_next_event, + .set_mode = lguest_clockevent_set_mode, + .rating = INT_MAX, + .mult = 1, + .shift = 0, + .min_delta_ns = LG_CLOCK_MIN_DELTA, + .max_delta_ns = LG_CLOCK_MAX_DELTA, +}; + +/* This is the Guest timer interrupt handler (hardware interrupt 0). We just + * call the clockevent infrastructure and it does whatever needs doing. */ static void lguest_time_irq(unsigned int irq, struct irq_desc *desc) { + unsigned long flags; + /* Check in case host TSC has changed rate. */ if (unlikely(tsc_khz != lguest_data.tsc_khz)) { tsc_khz = lguest_data.tsc_khz; @@ -368,8 +419,11 @@ static void lguest_time_irq(unsigned int __get_cpu_var(sc_data).cyc2ns_scale = (1000000 << CYC2NS_SCALE_FACTOR) / tsc_khz; } - do_timer(hcall(LHCALL_TIMER_READ, 0, 0, 0)); - update_process_times(user_mode_vm(get_irq_regs())); + + /* Don't interrupt us while this is running. */ + local_irq_save(flags); + lguest_clockevent.event_handler(&lguest_clockevent); + local_irq_restore(flags); } static void lguest_time_init(void) @@ -379,7 +433,10 @@ static void lguest_time_init(void) lguest_clock.mult = clocksource_khz2mult(tsc_khz, 22); clocksource_register(&lguest_clock); - hcall(LHCALL_TIMER_READ, 0, 0, 0); + /* We can't set cpumask in the initializer: damn C limitations! */ + lguest_clockevent.cpumask = cpumask_of_cpu(0); + clockevents_register_device(&lguest_clockevent); + enable_lguest_irq(0); } =================================================================== --- a/drivers/lguest/lguest_user.c +++ b/drivers/lguest/lguest_user.c @@ -139,6 +139,7 @@ static int initialize(struct file *file, setup_regs(lg->regs, args[2]); setup_guest_gdt(lg); + init_clockdev(lg); lg->tsk = current; get_task_struct(lg->tsk); lg->mm = get_task_mm(lg->tsk); @@ -200,6 +201,8 @@ static int close(struct inode *inode, st return 0; mutex_lock(&lguest_lock); + /* Cancels the hrtimer set via LHCALL_SET_CLOCKEVENT. */ + hrtimer_cancel(&lg->hrt); release_all_dma(lg); free_guest_pagetable(lg); put_task_struct(lg->tsk); =================================================================== --- a/include/linux/lguest.h +++ b/include/linux/lguest.h @@ -15,7 +15,7 @@ #define LHCALL_LOAD_IDT_ENTRY 6 #define LHCALL_SET_STACK 7 #define LHCALL_TS 8 -#define LHCALL_TIMER_READ 9 +#define LHCALL_SET_CLOCKEVENT 9 #define LHCALL_HALT 10 #define LHCALL_GET_WALLCLOCK 11 #define LHCALL_BIND_DMA 12 @@ -23,6 +23,9 @@ #define LHCALL_SET_PTE 14 #define LHCALL_SET_PMD 15 #define LHCALL_LOAD_TLS 16 + +#define LG_CLOCK_MIN_DELTA 100UL +#define LG_CLOCK_MAX_DELTA ULONG_MAX #define LGUEST_TRAP_ENTRY 0x1F - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/