kvmarm.lists.cs.columbia.edu archive mirror
 help / color / mirror / Atom feed
From: Auger Eric <eric.auger@redhat.com>
To: Jingyi Wang <wangjingyi11@huawei.com>,
	drjones@redhat.com, kvm@vger.kernel.org,
	kvmarm@lists.cs.columbia.edu
Cc: maz@kernel.org
Subject: Re: [kvm-unit-tests PATCH v2 8/8] arm64: microbench: Add vtimer latency test
Date: Fri, 3 Jul 2020 09:45:51 +0200	[thread overview]
Message-ID: <a4bcb265-90c9-d5f3-c353-85f876cd5472@redhat.com> (raw)
In-Reply-To: <5a43242d-2c0b-d8d1-b12d-7436e7d03e52@huawei.com>

Hi Jingyi,

On 7/3/20 9:41 AM, Jingyi Wang wrote:
> Hi Eric, Drew,
> 
> On 7/2/2020 9:36 PM, Auger Eric wrote:
>> Hi Jingyi,
>>
>> On 7/2/20 5:01 AM, Jingyi Wang wrote:
>>> Trigger PPIs by setting up a 10msec timer and test the latency.
>>
>> so for each iteration the accumulated valued is 10 ms + latency, right?
>> and what is printed at the end does include the accumulated periods.
>> Wouldn't it make sense to have a test->post() that substract this value.
>> You would need to store the actual number of iterations.
>>
>> Thanks
>>
>> Eric
> 
> That's right, the result indicates 10ms + latency, which is a 10msec
> timer actually costs. I think using the difference instead of the total
> time cost can be a little confusing here. Maybe we can use test->post()
> to get the latency and print an extra result in logs? Do you have any
> opinions on that?

for other interrupts you only print the latency, right? Here if I
understand correctly you use the timer to trigger the PPI but still you
care about the latency, hence my suggestion to only print the latency.

Thanks

Eric
> 
> Thanks,
> Jingyi
> 
>>>
>>> Signed-off-by: Jingyi Wang <wangjingyi11@huawei.com>
>>> ---
>>>   arm/micro-bench.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++-
>>>   1 file changed, 55 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/arm/micro-bench.c b/arm/micro-bench.c
>>> index 4c962b7..6822084 100644
>>> --- a/arm/micro-bench.c
>>> +++ b/arm/micro-bench.c
>>> @@ -23,8 +23,13 @@
>>>   #include <asm/gic-v3-its.h>
>>>     #define NTIMES (1U << 16)
>>> +#define NTIMES_MINOR (1U << 8)
>>>   #define MAX_NS (5 * 1000 * 1000 * 1000UL)
>>>   +#define IRQ_VTIMER        27
>>> +#define ARCH_TIMER_CTL_ENABLE    (1 << 0)
>>> +#define ARCH_TIMER_CTL_IMASK    (1 << 1)
>>> +
>>>   static u32 cntfrq;
>>>     static volatile bool irq_ready, irq_received;
>>> @@ -33,9 +38,16 @@ static void (*write_eoir)(u32 irqstat);
>>>     static void gic_irq_handler(struct pt_regs *regs)
>>>   {
>>> +    u32 irqstat = gic_read_iar();
>>>       irq_ready = false;
>>>       irq_received = true;
>>> -    gic_write_eoir(gic_read_iar());
>>> +    gic_write_eoir(irqstat);
>>> +
>>> +    if (irqstat == IRQ_VTIMER) {
>>> +        write_sysreg((ARCH_TIMER_CTL_IMASK | ARCH_TIMER_CTL_ENABLE),
>>> +                 cntv_ctl_el0);
>>> +        isb();
>>> +    }
>>>       irq_ready = true;
>>>   }
>>>   @@ -189,6 +201,47 @@ static void lpi_exec(void)
>>>       assert_msg(irq_received, "failed to receive LPI in time, but
>>> received %d successfully\n", received);
>>>   }
>>>   +static bool timer_prep(void)
>>> +{
>>> +    static void *gic_isenabler;
>>> +
>>> +    gic_enable_defaults();
>>> +    install_irq_handler(EL1H_IRQ, gic_irq_handler);
>>> +    local_irq_enable();
>>> +
>>> +    gic_isenabler = gicv3_sgi_base() + GICR_ISENABLER0;
>>> +    writel(1 << 27, gic_isenabler);
>>> +    write_sysreg(ARCH_TIMER_CTL_ENABLE, cntv_ctl_el0);
>>> +    isb();
>>> +
>>> +    gic_prep_common();
>>> +    return true;
>>> +}
>>> +
>>> +static void timer_exec(void)
>>> +{
>>> +    u64 before_timer;
>>> +    u64 timer_10ms;
>>> +    unsigned tries = 1 << 28;
>>> +    static int received = 0;
>>> +
>>> +    irq_received = false;
>>> +
>>> +    before_timer = read_sysreg(cntvct_el0);
>>> +    timer_10ms = cntfrq / 100;
>>> +    write_sysreg(before_timer + timer_10ms, cntv_cval_el0);
>>> +    write_sysreg(ARCH_TIMER_CTL_ENABLE, cntv_ctl_el0);
>>> +    isb();
>>> +
>>> +    while (!irq_received && tries--)
>>> +        cpu_relax();
>>> +
>>> +    if (irq_received)
>>> +        ++received;
>>> +
>>> +    assert_msg(irq_received, "failed to receive PPI in time, but
>>> received %d successfully\n", received);
>>> +}
>>> +
>>>   static void hvc_exec(void)
>>>   {
>>>       asm volatile("mov w0, #0x4b000000; hvc #0" ::: "w0");
>>> @@ -236,6 +289,7 @@ static struct exit_test tests[] = {
>>>       {"ipi",            ipi_prep,    ipi_exec,        NTIMES,       
>>> true},
>>>       {"ipi_hw",        ipi_hw_prep,    ipi_exec,       
>>> NTIMES,        true},
>>>       {"lpi",            lpi_prep,    lpi_exec,        NTIMES,       
>>> true},
>>> +    {"timer_10ms",        timer_prep,    timer_exec,       
>>> NTIMES_MINOR,    true},
>>>   };
>>>     struct ns_time {
>>>
>>
>>
>> .
>>
> 

_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

  reply	other threads:[~2020-07-03  7:46 UTC|newest]

Thread overview: 39+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-02  3:01 [kvm-unit-tests PATCH v2 0/8] arm/arm64: Add IPI/LPI/vtimer latency test Jingyi Wang
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 1/8] arm64: microbench: get correct ipi received num Jingyi Wang
2020-07-02 12:36   ` Auger Eric
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 2/8] arm64: microbench: Use the funcions for ipi test as the general functions for gic(ipi/lpi/timer) test Jingyi Wang
2020-07-02  5:25   ` Andrew Jones
2020-07-02  8:21     ` Jingyi Wang
2020-07-02 12:36   ` Auger Eric
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 3/8] arm64: microbench: gic: Add gicv4.1 support for ipi latency test Jingyi Wang
2020-07-02  8:22   ` Marc Zyngier
2020-07-02  9:02     ` Jingyi Wang
2020-07-02  9:17       ` Marc Zyngier
2020-07-02  9:29         ` Jingyi Wang
2020-07-02 12:36     ` Auger Eric
2020-07-02 13:03       ` Marc Zyngier
2020-07-02 12:57   ` Auger Eric
2020-07-02 13:08     ` Marc Zyngier
2020-07-02 13:42       ` Auger Eric
2020-07-03  3:39         ` Jingyi Wang
2020-07-02 21:33     ` Andrew Jones
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 4/8] arm64: its: Handle its command queue wrapping Jingyi Wang
2020-07-02 13:01   ` Auger Eric
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 5/8] arm64: microbench: its: Add LPI latency test Jingyi Wang
2020-07-02 13:13   ` Auger Eric
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 6/8] arm64: microbench: Allow each test to specify its running times Jingyi Wang
2020-07-02  5:29   ` Andrew Jones
2020-07-02  8:46     ` Jingyi Wang
2020-07-02 13:17     ` Auger Eric
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 7/8] arm64: microbench: Add time limit for each individual test Jingyi Wang
2020-07-02  5:48   ` Andrew Jones
2020-07-02  8:47     ` Jingyi Wang
2020-07-02 13:23   ` Auger Eric
2020-07-03  3:42     ` Jingyi Wang
2020-07-02  3:01 ` [kvm-unit-tests PATCH v2 8/8] arm64: microbench: Add vtimer latency test Jingyi Wang
2020-07-02  5:44   ` Andrew Jones
2020-07-02  8:56     ` Jingyi Wang
2020-07-02 13:36   ` Auger Eric
2020-07-03  7:41     ` Jingyi Wang
2020-07-03  7:45       ` Auger Eric [this message]
2020-07-06 12:23         ` Jingyi Wang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=a4bcb265-90c9-d5f3-c353-85f876cd5472@redhat.com \
    --to=eric.auger@redhat.com \
    --cc=drjones@redhat.com \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=maz@kernel.org \
    --cc=wangjingyi11@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).