linux-kernel - Re: [PATCH v4 07/10] KVM: selftests: Support multiple vCPUs in demand paging test

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: <20200124101637.4ppqzq7fxrd4ospw@kamzik.brq.redhat.com>
Date:   Fri, 24 Jan 2020 11:16:37 +0100
From:   Andrew Jones <drjones@...hat.com>
To:     Ben Gardon <bgardon@...gle.com>
Cc:     linux-kernel@...r.kernel.org, kvm@...r.kernel.org,
        linux-kselftest@...r.kernel.org,
        Paolo Bonzini <pbonzini@...hat.com>,
        Cannon Matthews <cannonmatthews@...gle.com>,
        Peter Xu <peterx@...hat.com>, Peter Shier <pshier@...gle.com>,
        Oliver Upton <oupton@...gle.com>
Subject: Re: [PATCH v4 07/10] KVM: selftests: Support multiple vCPUs in
 demand paging test

On Thu, Jan 23, 2020 at 10:04:33AM -0800, Ben Gardon wrote:
> Most VMs have multiple vCPUs, the concurrent execution of which has a
> substantial impact on demand paging performance. Add an option to create
> multiple vCPUs to each access disjoint regions of memory.
> 
> Signed-off-by: Ben Gardon <bgardon@...gle.com>
> ---
>  .../selftests/kvm/demand_paging_test.c        | 255 ++++++++++++------
>  1 file changed, 172 insertions(+), 83 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/demand_paging_test.c b/tools/testing/selftests/kvm/demand_paging_test.c
> index 9e2a5f7dfa140..2002032df32cc 100644
> --- a/tools/testing/selftests/kvm/demand_paging_test.c
> +++ b/tools/testing/selftests/kvm/demand_paging_test.c
> @@ -24,8 +24,6 @@
>  #include "kvm_util.h"
>  #include "processor.h"
>  
> -#define VCPU_ID				1
> -
>  /* The memory slot index demand page */
>  #define TEST_MEM_SLOT_INDEX		1
>  
> @@ -34,6 +32,14 @@
>  
>  #define DEFAULT_GUEST_TEST_MEM_SIZE (1 << 30) /* 1G */
>  
> +#ifdef PRINT_PER_VCPU_UPDATES
> +#define PER_VCPU_DEBUG(...) DEBUG(__VA_ARGS__)
> +#else
> +#define PER_VCPU_DEBUG(...)
> +#endif
> +
> +#define MAX_VCPUS 512
> +
>  /*
>   * Guest/Host shared variables. Ensure addr_gva2hva() and/or
>   * sync_global_to/from_guest() are used when accessing from
> @@ -67,18 +73,25 @@ struct vcpu_args {
>  	struct kvm_vm *vm;
>  };
>  
> -static struct vcpu_args vcpu_args;
> +static struct vcpu_args vcpu_args[MAX_VCPUS];
>  
>  /*
>   * Continuously write to the first 8 bytes of each page in the demand paging
>   * memory region.
>   */
> -static void guest_code(void)
> +static void guest_code(uint32_t vcpu_id)
>  {
> -	uint64_t gva = vcpu_args.gva;
> -	uint64_t pages = vcpu_args.pages;
> +	uint64_t gva;
> +	uint64_t pages;
>  	int i;
>  
> +	/* Return to signal error if vCPU args data structure is courrupt. */
> +	if (vcpu_args[vcpu_id].vcpu_id != vcpu_id)
> +		return;

This should be GUEST_ASSERT(vcpu_args[vcpu_id].vcpu_id == vcpu_id), which
will do a UCALL_ABORT when it fails. Otherwise we're returning to where?
Likely we'll get an exception of some sort when we return to nothing,
but that's not a very clean way to die.

Thanks,
drew