Re: [PATCH v13 10/13] x86/sgx: Add sgx_einit() for initializing enclaves

[Huang, Kai]

On Mon, 2018-08-27 at 21:53 +0300, Jarkko Sakkinen wrote:
> From: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> 
> Add a function to perform ENCLS(EINIT), which initializes an enclave,
> which can be used by a driver for running enclaves and VMMs.
> 
> Writing the LE hash MSRs is extraordinarily expensive, e.g. 3-4x
> slower
> than normal MSRs, so we use a per-cpu cache to track the last known
> value
> of the MSRs to avoid unnecessarily writing the MSRs with the current
> value.
> 
> Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> Co-developed-by: Jarkko Sakkinen <jarkko.sakkinen@xxxxxxxxxxxxxxx>
> Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@xxxxxxxxxxxxxxx>
> ---
>  arch/x86/include/asm/sgx.h      |  2 +
>  arch/x86/kernel/cpu/intel_sgx.c | 86
> +++++++++++++++++++++++++++++++--
>  2 files changed, 85 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/include/asm/sgx.h b/arch/x86/include/asm/sgx.h
> index baf30d49b71f..c15c156436be 100644
> --- a/arch/x86/include/asm/sgx.h
> +++ b/arch/x86/include/asm/sgx.h
> @@ -108,6 +108,8 @@ void sgx_free_page(struct sgx_epc_page *page);
>  void sgx_page_reclaimable(struct sgx_epc_page *page);
>  struct page *sgx_get_backing(struct file *file, pgoff_t index);
>  void sgx_put_backing(struct page *backing_page, bool write);
> +int sgx_einit(struct sgx_sigstruct *sigstruct, struct sgx_einittoken
> *token,
> +	      struct sgx_epc_page *secs_page, u64 lepubkeyhash[4]);
>  
>  #define ENCLS_FAULT_FLAG 0x40000000UL
>  #define ENCLS_FAULT_FLAG_ASM "$0x40000000"
> diff --git a/arch/x86/kernel/cpu/intel_sgx.c
> b/arch/x86/kernel/cpu/intel_sgx.c
> index 1046478a3ab9..fe25e6805680 100644
> --- a/arch/x86/kernel/cpu/intel_sgx.c
> +++ b/arch/x86/kernel/cpu/intel_sgx.c
> @@ -9,6 +9,7 @@
>  #include <linux/sched/signal.h>
>  #include <linux/shmem_fs.h>
>  #include <linux/slab.h>
> +#include <linux/suspend.h>
>  #include <asm/sgx.h>
>  #include <asm/sgx_pr.h>
>  
> @@ -38,6 +39,18 @@ static LIST_HEAD(sgx_active_page_list);
>  static DEFINE_SPINLOCK(sgx_active_page_list_lock);
>  static struct task_struct *ksgxswapd_tsk;
>  static DECLARE_WAIT_QUEUE_HEAD(ksgxswapd_waitq);
> +static struct notifier_block sgx_pm_notifier;
> +static u64 sgx_pm_cnt;
> +
> +/* The cache for the last known values of IA32_SGXLEPUBKEYHASHx MSRs
> for each
> + * CPU. The entries are initialized when they are first used by
> sgx_einit().
> + */
> +struct sgx_lepubkeyhash {
> +	u64 msrs[4];
> +	u64 pm_cnt;

May I ask why do we need pm_cnt here? In fact why do we need suspend
staff (namely, sgx_pm_cnt above, and related code in this patch) here
in this patch? From the patch commit message I don't see why we need PM
staff here. Please give comment why you need PM staff, or you may
consider to split the PM staff to another patch.

> +};
> +
> +static DEFINE_PER_CPU(struct sgx_lepubkeyhash *,
> sgx_lepubkeyhash_cache);
>  
>  /**
>   * sgx_reclaim_pages - reclaim EPC pages from the consumers
> @@ -328,6 +341,54 @@ void sgx_put_backing(struct page *backing_page,
> bool write)
>  }
>  EXPORT_SYMBOL_GPL(sgx_put_backing);
>  
> +/**
> + * sgx_einit - initialize an enclave
> + * @sigstruct:		a pointer to the SIGSTRUCT
> + * @token:		a pointer to the EINITTOKEN
> + * @secs_page:		a pointer to the SECS EPC page
> + * @lepubkeyhash:	the desired value for IA32_SGXLEPUBKEYHASHx
> MSRs
> + *
> + * Try to perform EINIT operation. If the MSRs are writable, they
> are updated
> + * according to @lepubkeyhash.
> + *
> + * Return:
> + *   0 on success,
> + *   -errno on failure
> + *   SGX error code if EINIT fails
> + */
> +int sgx_einit(struct sgx_sigstruct *sigstruct, struct sgx_einittoken
> *token,
> +	      struct sgx_epc_page *secs_page, u64 lepubkeyhash[4])
> +{
> +	struct sgx_lepubkeyhash __percpu *cache;
> +	bool cache_valid;
> +	int i, ret;
> +
> +	if (!sgx_lc_enabled)
> +		return __einit(sigstruct, token,
> sgx_epc_addr(secs_page));
> +
> +	cache = per_cpu(sgx_lepubkeyhash_cache, smp_processor_id());
> +	if (!cache) {
> +		cache = kzalloc(sizeof(struct sgx_lepubkeyhash),
> GFP_KERNEL);
> +		if (!cache)
> +			return -ENOMEM;
> +	}

It seems per-cpu variable is a pointer to struct sgx_lepubkeyhash, and
the actual structure is allocated at the first time the function is
called. May I ask when will it be freed? It seems the free is not in
this patch. Or I am misunderstanding something?

> +
> +	cache_valid = cache->pm_cnt == sgx_pm_cnt;
> +	cache->pm_cnt = sgx_pm_cnt;
> +	preempt_disable();
> +	for (i = 0; i < 4; i++) {
> +		if (cache_valid && lepubkeyhash[i] == cache-
> >msrs[i])
> +			continue;
> +
> +		wrmsrl(MSR_IA32_SGXLEPUBKEYHASH0 + i,
> lepubkeyhash[i]);
> +		cache->msrs[i] = lepubkeyhash[i];
> +	}
> +	ret = __einit(sigstruct, token, sgx_epc_addr(secs_page));
> +	preempt_enable();
> +	return ret;
> +}
> +EXPORT_SYMBOL(sgx_einit);
> +
>  static __init int sgx_init_epc_bank(u64 addr, u64 size, unsigned
> long index,
>  				    struct sgx_epc_bank *bank)
>  {
> @@ -426,6 +487,15 @@ static __init int sgx_page_cache_init(void)
>  	return 0;
>  }
>  
> +static int sgx_pm_notifier_cb(struct notifier_block *nb, unsigned
> long action,
> +			      void *data)
> +{
> +	if (action == PM_SUSPEND_PREPARE || action ==
> PM_HIBERNATION_PREPARE)
> +		sgx_pm_cnt++;
> +
> +	return NOTIFY_DONE;
> +}
> +
>  static __init int sgx_init(void)
>  {
>  	struct task_struct *tsk;
> @@ -452,20 +522,30 @@ static __init int sgx_init(void)
>  	if (!(fc & FEATURE_CONTROL_SGX_LE_WR))
>  		pr_info("IA32_SGXLEPUBKEYHASHn MSRs are not
> writable\n");
>  
> -	ret = sgx_page_cache_init();
> +	sgx_pm_notifier.notifier_call = sgx_pm_notifier_cb;
> +	ret = register_pm_notifier(&sgx_pm_notifier);
>  	if (ret)
>  		return ret;
>  
> +	ret = sgx_page_cache_init();
> +	if (ret)
> +		goto out_pm;
> +
>  	tsk = kthread_run(ksgxswapd, NULL, "ksgxswapd");
>  	if (IS_ERR(tsk)) {
> -		sgx_page_cache_teardown();
> -		return PTR_ERR(tsk);
> +		ret = PTR_ERR(tsk);
> +		goto out_pcache;
>  	}
>  	ksgxswapd_tsk = tsk;
>  
>  	sgx_enabled = true;
>  	sgx_lc_enabled = !!(fc & FEATURE_CONTROL_SGX_LE_WR);
>  	return 0;
> +out_pcache:
> +	sgx_page_cache_teardown();

I don't think this particular 2 lines of code of 'out_pcache' case
should be in this patch?

Thanks,
-Kai

> +out_pm:
> +	unregister_pm_notifier(&sgx_pm_notifier);
> +	return ret;
>  }
>  
>  arch_initcall(sgx_init);