Sean Christopherson <seanjc@xxxxxxxxxx> writes:
> On Mon, Nov 22, 2021, Vitaly Kuznetsov wrote:
>> With the elevated 'KVM_CAP_MAX_VCPUS' value kvm_create_max_vcpus test
>> may hit RLIMIT_NOFILE limits:
>>
>> # ./kvm_create_max_vcpus
>> KVM_CAP_MAX_VCPU_ID: 4096
>> KVM_CAP_MAX_VCPUS: 1024
>> Testing creating 1024 vCPUs, with IDs 0...1023.
>> /dev/kvm not available (errno: 24), skipping test
>>
>> Adjust RLIMIT_NOFILE limits to make sure KVM_CAP_MAX_VCPUS fds can be
>> opened. Note, raising hard limit ('rlim_max') requires CAP_SYS_RESOURCE
>> capability which is generally not needed to run kvm selftests (but without
>> raising the limit the test is doomed to fail anyway).
>>
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@xxxxxxxxxx>
>> ---
>> .../selftests/kvm/kvm_create_max_vcpus.c | 22 +++++++++++++++++++
>> 1 file changed, 22 insertions(+)
>>
>> diff --git a/tools/testing/selftests/kvm/kvm_create_max_vcpus.c b/tools/testing/selftests/kvm/kvm_create_max_vcpus.c
>> index f968dfd4ee88..19198477a10e 100644
>> --- a/tools/testing/selftests/kvm/kvm_create_max_vcpus.c
>> +++ b/tools/testing/selftests/kvm/kvm_create_max_vcpus.c
>> @@ -12,6 +12,7 @@
>> #include <stdio.h>
>> #include <stdlib.h>
>> #include <string.h>
>> +#include <sys/resource.h>
>>
>> #include "test_util.h"
>>
>> @@ -19,6 +20,9 @@
>> #include "asm/kvm.h"
>> #include "linux/kvm.h"
>>
>> +/* 'Safe' number of open file descriptors in addition to vCPU fds needed */
>> +#define NOFD 16
>
> Any reason not to make this "buffer" extra large, e.g. 100+ to avoid having to
> debug this issue again in the future?
>
No, not really. We could've avoided this ambiguity completely by
checking how many fds are already open but all methods I can think of
are 'too much'. In my testing I needed around 10 so I put '16' but '100'
is even better.
>> +
>> void test_vcpu_creation(int first_vcpu_id, int num_vcpus)
>> {
>> struct kvm_vm *vm;
>> @@ -40,10 +44,28 @@ int main(int argc, char *argv[])
>> {
>> int kvm_max_vcpu_id = kvm_check_cap(KVM_CAP_MAX_VCPU_ID);
>> int kvm_max_vcpus = kvm_check_cap(KVM_CAP_MAX_VCPUS);
>
> Rather than a separate define that's hard to describe succintly, what about:
>
> int nr_fds_wanted = kvm_max_vcpus + <arbitrary number>
>
> and then the body becomes
>
> if (nr_fds_wanted > rl.rlim_cur) {
> rl.rlim_cur = nr_fds_wanted;
> rl.rlim_max = max(rl.rlim_max, nr_fds_wanted);
>
> ...
> }
Sure but a "succinct" comment will still be needed, either near the
'NOFD' define or above 'int nr_fds_wanted' :-)
>
>> + struct rlimit rl;
>>
>> pr_info("KVM_CAP_MAX_VCPU_ID: %d\n", kvm_max_vcpu_id);
>> pr_info("KVM_CAP_MAX_VCPUS: %d\n", kvm_max_vcpus);
>>
>> + /*
>> + * Creating KVM_CAP_MAX_VCPUS vCPUs require KVM_CAP_MAX_VCPUS open
>> + * file decriptors.
>> + */
>> + TEST_ASSERT(!getrlimit(RLIMIT_NOFILE, &rl),
>> + "getrlimit() failed (errno: %d)", errno);
>
> And strerror() output too?
>
Sure, will add in v2.
>> +
>> + if (kvm_max_vcpus > rl.rlim_cur - NOFD) {
>> + rl.rlim_cur = kvm_max_vcpus + NOFD;
>> +
>> + if (kvm_max_vcpus > rl.rlim_max - NOFD)
>> + rl.rlim_max = kvm_max_vcpus + NOFD;
>> +
>> + TEST_ASSERT(!setrlimit(RLIMIT_NOFILE, &rl),
>> + "setrlimit() failed (errno: %d)", errno);
>> + }
>> +
>> /*
>> * Upstream KVM prior to 4.8 does not support KVM_CAP_MAX_VCPU_ID.
>> * Userspace is supposed to use KVM_CAP_MAX_VCPUS as the maximum ID
>> --
>> 2.33.1
>>
>
--
Vitaly
