Re: [PATCH v6 2/2] x86/sgx: Add an attribute for the amount of SGX memory in a NUMA node

[Greg Kroah-Hartman]

On Tue, Sep 28, 2021 at 06:13:50AM +0300, Jarkko Sakkinen wrote:
> The amount of SGX memory on the system is determined by the BIOS and it
> varies wildly between systems.  It can be from dozens of MB's on desktops
> or VM's, up to many GB's on servers.  Just like for regular memory, it is
> sometimes useful to know the amount of usable SGX memory in the system.
> 
> Add an attribute for the amount of SGX memory in bytes to each NUMA
> node. The path is /sys/devices/system/node/node[0-9]*/sgx/memory_size.
> Calculate these values by summing up EPC section sizes for each node
> during the driver initalization.
> 
> Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> Signed-off-by: Jarkko Sakkinen <jarkko@xxxxxxxxxx>
> ---
> 
> v6:
> * Initialize node->size to zero in sgx_setup_epc_section(), when the
>   node is first accessed. The bug report:
>   https://lore.kernel.org/linux-sgx/f45245ba-41b8-62ae-38b5-64725a214bad@xxxxxxxxx/
> 
> v5:
> * A new patch based on the discussion on
>   https://lore.kernel.org/linux-sgx/3a7cab4115b4f902f3509ad8652e616b91703e1d.camel@xxxxxxxxxx/T/#t
> 
>  Documentation/x86/sgx.rst      | 14 ++++++

sysfs files have to be documented in Documentation/ABI/ so that they can
be automatically checked, and added to the documentation output
properly.  Please do that here as well.


>  arch/x86/kernel/cpu/sgx/main.c | 91 ++++++++++++++++++++++++++++++++++
>  arch/x86/kernel/cpu/sgx/sgx.h  |  2 +
>  3 files changed, 107 insertions(+)
> 
> diff --git a/Documentation/x86/sgx.rst b/Documentation/x86/sgx.rst
> index dd0ac96ff9ef..f9d9cfa6dbf9 100644
> --- a/Documentation/x86/sgx.rst
> +++ b/Documentation/x86/sgx.rst
> @@ -250,3 +250,17 @@ user wants to deploy SGX applications both on the host and in guests
>  on the same machine, the user should reserve enough EPC (by taking out
>  total virtual EPC size of all SGX VMs from the physical EPC size) for
>  host SGX applications so they can run with acceptable performance.
> +
> +Per NUMA node SGX attributes
> +============================
> +
> +NUMA nodes devices expose SGX specific attributes in the following path:
> +
> +	/sys/devices/system/node/node[0-9]*/sgx/
> +
> +Attributes
> +----------
> +
> +memory_size
> +                Total available physical SGX memory, also known as Enclave
> +                Page Cache (EPC), in bytes.
> diff --git a/arch/x86/kernel/cpu/sgx/main.c b/arch/x86/kernel/cpu/sgx/main.c
> index a6e313f1a82d..4f1e3b5e3d14 100644
> --- a/arch/x86/kernel/cpu/sgx/main.c
> +++ b/arch/x86/kernel/cpu/sgx/main.c
> @@ -714,9 +714,11 @@ static bool __init sgx_page_cache_init(void)
>  			spin_lock_init(&sgx_numa_nodes[nid].lock);
>  			INIT_LIST_HEAD(&sgx_numa_nodes[nid].free_page_list);
>  			node_set(nid, sgx_numa_mask);
> +			sgx_numa_nodes[nid].size = 0;
>  		}
>  
>  		sgx_epc_sections[i].node =  &sgx_numa_nodes[nid];
> +		sgx_numa_nodes[nid].size += size;
>  
>  		sgx_nr_epc_sections++;
>  	}
> @@ -790,6 +792,87 @@ int sgx_set_attribute(unsigned long *allowed_attributes,
>  }
>  EXPORT_SYMBOL_GPL(sgx_set_attribute);
>  
> +#ifdef CONFIG_NUMA
> +static void sgx_numa_exit(void)
> +{
> +	int nid;
> +
> +	for (nid = 0; nid < num_possible_nodes(); nid++) {
> +		if (!sgx_numa_nodes[nid].kobj)
> +			continue;
> +
> +		kobject_put(sgx_numa_nodes[nid].kobj);
> +	}
> +}
> +
> +#define SGX_NODE_ATTR_RO(_name) \
> +	static struct kobj_attribute _name##_attr = __ATTR_RO(_name)

Why are you dealing with a "raw" kobject?  Shouldn't you have a device
and use a device attribute?

> +static bool sgx_numa_init(void)
> +{
> +	struct sgx_numa_node *node;
> +	struct device *dev;
> +	int nid;
> +	int ret;
> +
> +	for (nid = 0; nid < num_possible_nodes(); nid++) {
> +		if (!sgx_numa_nodes[nid].size)
> +			continue;
> +
> +		node = &sgx_numa_nodes[nid];
> +		dev = &node_devices[nid]->dev;
> +
> +		node->kobj = kobject_create_and_add("sgx", &dev->kobj);

You just "broke" the tree by putting a raw kobject below a struct
device.  Please do not do that.

> +		if (!node->kobj) {
> +			sgx_numa_exit();
> +			return false;
> +		}
> +
> +		ret = sysfs_create_group(node->kobj, &sgx_node_attr_group);

And you raced with userspace and lost.

Wait, you have a kobject _just_ for a subdirectory name?  Why?  Use a
named attribute group, that's exactly what that is for.

Properly attach your attributes to the device you have, don't do extra
work and complex code that you do not have to at all.

thanks,

greg k-h