Re: [PATCH 1/3] xen/domain: Reorder trivial initialisation in early domain_create()

From: Andrew Cooper <andrew.cooper3@citrix.com>
To: Jan Beulich <jbeulich@suse.com>
Cc: "Roger Pau Monné" <roger.pau@citrix.com>, "Wei Liu" <wl@xen.org>,
	"Stefano Stabellini" <sstabellini@kernel.org>,
	"Julien Grall" <julien@xen.org>,
	"Volodymyr Babchuk" <Volodymyr_Babchuk@epam.com>,
	Xen-devel <xen-devel@lists.xenproject.org>
Subject: Re: [PATCH 1/3] xen/domain: Reorder trivial initialisation in early domain_create()
Date: Tue, 22 Dec 2020 10:24:37 +0000	[thread overview]
Message-ID: <472745b0-7b9b-1412-85c7-6186711fadd8@citrix.com> (raw)
In-Reply-To: <3397707d-ba05-4db2-7dfd-e18dbe044a26@suse.com>

On 22/12/2020 10:10, Jan Beulich wrote:
> On 21.12.2020 19:14, Andrew Cooper wrote:
>> --- a/xen/common/domain.c
>> +++ b/xen/common/domain.c
>> @@ -391,25 +391,7 @@ struct domain *domain_create(domid_t domid,
>>  
>>      TRACE_1D(TRC_DOM0_DOM_ADD, d->domain_id);
>>  
>> -    /*
>> -     * Allocate d->vcpu[] and set ->max_vcpus up early.  Various per-domain
>> -     * resources want to be sized based on max_vcpus.
>> -     */
>> -    if ( !is_system_domain(d) )
>> -    {
>> -        err = -ENOMEM;
>> -        d->vcpu = xzalloc_array(struct vcpu *, config->max_vcpus);
>> -        if ( !d->vcpu )
>> -            goto fail;
>> -
>> -        d->max_vcpus = config->max_vcpus;
>> -    }
>> -
>> -    lock_profile_register_struct(LOCKPROF_TYPE_PERDOM, d, domid);
> Wouldn't this also count as "trivial initialization", and hence while
> moving want to at least be placed ...
>
>> -    if ( (err = xsm_alloc_security_domain(d)) != 0 )
>> -        goto fail;
>> -
>> +    /* Trivial initialisation. */
>>      atomic_set(&d->refcnt, 1);
>>      RCU_READ_LOCK_INIT(&d->rcu_lock);
>>      spin_lock_init_prof(d, domain_lock);
>> @@ -434,6 +416,27 @@ struct domain *domain_create(domid_t domid,
>>      INIT_LIST_HEAD(&d->pdev_list);
>>  #endif
>>  
>> +    /* All error paths can depend on the above setup. */
> ... ahead of this comment?

Can do.

>
>> +    /*
>> +     * Allocate d->vcpu[] and set ->max_vcpus up early.  Various per-domain
>> +     * resources want to be sized based on max_vcpus.
>> +     */
>> +    if ( !is_system_domain(d) )
>> +    {
>> +        err = -ENOMEM;
>> +        d->vcpu = xzalloc_array(struct vcpu *, config->max_vcpus);
>> +        if ( !d->vcpu )
>> +            goto fail;
>> +
>> +        d->max_vcpus = config->max_vcpus;
>> +    }
>> +
>> +    lock_profile_register_struct(LOCKPROF_TYPE_PERDOM, d, domid);
>> +
>> +    if ( (err = xsm_alloc_security_domain(d)) != 0 )
>> +        goto fail;
>> +
>>      err = -ENOMEM;
>>      if ( !zalloc_cpumask_var(&d->dirty_cpumask) )
>>          goto fail;
>>
> Just as an observation (i.e. unlikely for this patch) I doubt
> system domains need ->dirty_cpumask set up, and hence this
> allocation may also want moving down a few lines.

I agree in principle.  However, something does (or at least did)
reference this for system domains when I did the is_system_domain() cleanup.

The fix might not be as trivial as just moving the allocation.

~Andrew